Business Technology Roundtable

Applied-AI adoption inside enterprises has nearly tripled in four years, moving from roughly 7 percent of OECD firms in 2021 to 20 percent in 2025.Over the same stretch of time, the share of the workforce demonstrating recognized AI skills has barely cleared 1 percent.That gap, not the pace of AI model releases, is the strategic problem every C-suite now has to own and purposefully resolve.The Widening Gap Between Deployment and CapabilityHalf of small and medium enterprises now cite a lack of skills as the reason they have not adopted generative AI tools, according to the OECD's newly declassified paper, Skills in the AI Age.Across 10 member countries, roughly one third of job vacancies already sit in occupations with high AI exposure, a share expected to keep climbing as generative tools embed themselves in everyday workplace software.Executives who treat this as a training line item rather than a workforce architecture problem will spend the next few years chasing a target that keeps moving faster than their budget cycle.Estonia has The Closest Thing to a National ResponseOf the dozens of national programs the OECD surveyed, Estonia's AI Leap initiative stands out as the most complete answer to the pace mismatch.Launched in 2025 as a public-private partnership with OpenAI, Anthropic, and domestic technology firms, it has already reached more than 20,000 high school students and 3,000 teachers with AI tool use and critical thinking instruction.A second phase beginning in 2026 extends the same curriculum to 38,000 vocational students and 2,000 instructors, with free laptops for low-income households and rural access built into the design rather than bolted on afterward.That combination, government sponsorship, direct involvement from the labs building the underlying technology, and equity provisions for the students least likely to be reached otherwise, is what separates a genuine workforce strategy from a training budget line.Finland and Korea Round Out the PictureFinland's Elements of AI course, built with the University of Helsinki and translated into multiple languages, remains the most widely diffused AI literacy model in the OECD sample and predates most of the current wave of national initiatives.Korea's AID 30+ project takes a different angle, targeting adults already in the workforce with annual vouchers worth up to KRW 350,000, a network of 100 designated universities, and integration into the national Credit Bank System so re-skilling actually counts toward a credential.Together, these three countries cover the full pipeline: general literacy, systemic preparation for students entering the workforce, and financed re-skilling for adults already in it.What This Means for Executive TeamsNone of this removes the responsibility from employers. The OECD is direct in noting that employer-led training remains the dominant form of adult learning in every member economy, and that firms routinely under-invest because they fear losing trained employees to competitors.Germany's wage-subsidy model, which can cover all training costs and up to 100 percent of wage costs for low-skilled workers, offers a template worth studying regardless of jurisdiction.The organizations that treat national skills infrastructure as a resource to plug into, rather than a substitute for their own re-skilling budget, are the ones that will actually close the gap before the next AI model generation makes today's skill set obsolete.The technology will keep moving faster than most organizations can retrain for it. The real competitive question heading into 2027 is not which AI model your company adopts next, but whether your workforce architecture was built to keep pace with the one after that.Reach out to learn more about our Applied-AI Initiative objectives.More...

A strategic business case and global market assessmentThe enterprise artificial intelligence landscape is undergoing a profound structural transition, particularly with the Professional Services related industries.As AI foundation models commoditize basic text generation and public datasets reach cognitive density limits, forward-thinking organizations are realizing that generic intelligence offers no sustainable competitive advantage.In response, enterprise architects and strategic advisors are shifting focus toward architectures that capture, structure, and monetize internal institutional knowledge.Central to this structural pivot is the concept of the Proprietary Intelligence Foundry.The foundry represents a strategic pipeline framework designed to transform an enterprise's undocumented human judgment into a permanent, defensible, and sovereign digital asset.Conceptual FoundationsThe Proprietary Intelligence Foundry is defined as a strategic pipeline architecture that captures an organization's tacit knowledge and daily operational exhaust, converting it into a governed knowledge graph to train or fine-tune specialized artificial intelligence models. Developed to resolve the inherent limitations of mainstream enterprise artificial intelligence implementations, the framework addresses a fundamental structural flaw in early generative deployment: the reliance on generic models paired with flat retrieval mechanisms.Limitations of Standard Retrieval-Augmented GenerationEarly large enterprise artificial intelligence deployments relied heavily on standard Retrieval-Augmented Generation (RAG) architectures.In a conventional RAG implementation, an enterprise indexes unstructured documents into a flat vector database using embedding models, subsequently retrieving semantic matches to populate a large language model prompt.While effective for basic information retrieval, standard RAG fails to capture the cognitive context, underlying heuristics, or decision-making logic of the organization's human experts.Standard RAG operates exclusively on explicit, written artifacts. However, an enterprise's core competitive advantage rarely resides in static documents; it lives in the dynamic, unwritten reasoning behind business decisions. Standard vector search lacks awareness of business relationships, temporal commitments, regulatory constraints, and institutional precedent.Consequently, RAG-driven systems frequently produce responses that are factually present in the underlying text but strategically naive.The Pipeline: Operational Exhaust to Governed Knowledge GraphThe Proprietary Intelligence Foundry replaces flat retrieval architectures with a continuous, structured pipeline designed to capture intelligence at its point of origin.The pipeline transitions raw operational activities into actionable reasoning systems through three interconnected stages:First, the system passively extracts signals from daily work artifacts, capturing operational exhaust such as email streams, client negotiation threads, video conference transcripts, collaboration patterns, and document revision histories.By capturing work exhaust ambiently, the system bypasses the need for domain experts to manually annotate data or interrupt billable workflows.Second, before any model retrieval or inference occurs, this captured exhaust is ingested into a governed knowledge graph mapped against domain-specific ontologies. The knowledge graph explicitly transforms unstructured interactions into a network of entities, obligations, decisions, and relationships.This relational mapping defines why a decision was reached, who authorized it, what trade-offs were accepted, and how operational exceptions were resolved.Third, the governed knowledge graph functions as an auditable training curriculum. Rather than relying exclusively on massive, multi-billion-parameter frontier models hosted by external third parties, the enterprise utilizes its knowledge graph to fine-tune smaller, highly specialized language models (SLMs) or agentic systems.These purpose-built AI models reason directly over the firm's specific judgment rather than averaging across public internet data.The Core Thesis: The Graph as the Durable AssetA central thesis is that in an environment of rapid foundational model development, the underlying language model functions as an ephemeral runtime layer. Public foundation models are continually updated, superseded, or open-sourced, rendering direct capital investments in model parameters subject to rapid economic depreciation.Conversely, the governed knowledge graph represents the durable, non-depreciating enterprise asset. By decoupling an organization's proprietary judgment from the underlying neural network runtime, the knowledge graph acts as a permanent substrate.As newer and more cost-effective model architectures emerge, the enterprise can systematically update the AI model runtime while keeping its proprietary knowledge layer fully intact.The conceptual architecture of the Proprietary Intelligence Foundry spans corporate governance, institutional memory, and the professional services intellectual capital defense.Harvesting Wisdom Versus Document PreservationTraditional knowledge management systems suffer from a historical focus on explicit document preservation — e.g. archiving static policy manuals, reports, and intranets.When experienced personnel retire or depart, organizations lose their tacit knowledge, including unwritten heuristics, pattern recognition, client intuition, and exception-handling capabilities built over decades.Tacit Knowledge Harvesting is the systematic extraction and structuring of this undocumented human experience. Applied-AI platforms allow enterprises to capture not merely final static work products, but the dynamic reasoning trajectories that produced them.By analyzing how top practitioners navigate complex scenarios, the Proprietary Intelligence Foundry codifies human reasoning into structured graph nodes and fine-tuning datasets, transforming ephemeral individual expertise into persistent organizational capital.Mitigating the Risk of Knowledge CollapseA critical strategic risk facing knowledge-intensive sectors — including legal services, strategy consulting, investment banking, and specialized engineering — is knowledge collapse.Knowledge collapse occurs when enterprises rely primarily on commercial, vendor-hosted frontier AI models trained on open, web-scale data.Because public foundation models generate responses based on statistical probabilities across public datasets, their outputs systematically regress toward the statistical center — representing the lowest common denominator of industry practice.If competing firms utilize identical AI vendor models, their operational workflows, client deliverables, and strategic insights inevitably converge.For professional service providers whose business models depend on charging premium rates for specialized judgment, adopting generic software tools risks eroding the distinct value proposition that justifies their fee structure.The Proprietary Intelligence Foundry eliminates knowledge collapse by anchoring intelligence generation strictly within the firm's unique cognitive history.Redefining Sovereign AI for the EnterpriseWhile national governments typically evaluate Sovereign AI through the lens of data residency laws and compute infrastructure, we can also redefine sovereignty for corporate leadership.Enterprise sovereignty hinges on a single essential question: Who owns the intelligence expressed through the Applied-AI system?An organization can host an open-weight model within a private cloud environment yet remain strategically non-sovereign if that model relies entirely on public internet-based reasoning.True enterprise sovereign intelligence requires total control over operational data and fine-tuning pipelines, the deployment of transparent, auditable small language models, absolute isolation from black-box vendor APIs to avoid token fee volatility, and the construction of a persistent digital capital asset reflecting the unique strategic DNA of the business.Macroeconomic Sizing and Market Demand AnalyticsQuantifying the total market opportunity for the Proprietary Intelligence Foundry requires synthesizing data across software, platform infrastructure, sovereign cloud, and vertical technology sectors.Because technology industry research firms currently do not track "firm-owned AI" as an isolated category, its addressable market must be evaluated by examining the broader categories it spans.Sizing Framework and Anchor DataThe demand for custom, proprietary intelligence architectures is anchored in macro-level enterprise capital reallocations.Data from Stanford University's Institute for Human-Centered Artificial Intelligence (HAI) indicates that global corporate investment in artificial intelligence reached $581.7 billion in 2025, a 130 percent year-over-year increase. Private artificial intelligence investment expanded to $344.7 billion (up 127.5 percent), with private funding for generative technologies growing more than 200 percent year-over-year.Within the United States, private AI investment reached $285.9 billion in 2025 — more than double the prior year's total, and roughly 23 times China's $12.4 billion over the same period.To contextualize the specific market pool for enterprise-owned intelligence stacks, published research forecasts from IDC, Gartner, McKinsey, and specialized market research firms establish the core market boundaries:Market SegmentPrimary Research SourceBaseline Sizing & YearForecast & Horizon YearCAGRWorldwide AI Software MarketIDC$64.0 Billion (2022)$251.0 Billion (2027)31.4% CAGREnterprise AI Solution Spend (Software + Services)IDC$307.0 Billion (2025)$632.0 Billion (2028)~27.2% CAGRAI Platforms and ModelsGartner$39.0 Billion (2025)$64.0 Billion (2026)63.4% Annual GrowthSovereign AI Infrastructure & Platforms (Proxy 1)MarketsandMarkets$40.0 Billion (2025)$148.0 Billion (2032)20.6% CAGRSovereign AI Infrastructure & Platforms (Proxy 2)Precedence Research$15.0 Billion (2025)$177.0 Billion (2035)28.0% CAGRSovereign AI Opportunity ScopeMcKinsey & Company$500–$600 Billion (By 2030)Total Addressable Opportunity—Legal AI Software Market (Vertical Proxy 1)MarketsandMarkets$3.11 Billion (2025)$10.82 Billion (2030)28.3% CAGRLegal AI Software Market (Vertical Proxy 2)Grand View Research$1.45 Billion (2024)$3.90 Billion (2030)~17.3% CAGR Analysis of Proxy Discrepancies and Market GrowthSynthesizing these market data points reveals critical trends regarding enterprise software adoption:First, platform spend is significantly outstripping general software growth. Gartner's "AI Platforms and Models" category — the closest segment tracking custom model customization and hosting infrastructure — is projected to grow 63.4 percent in a single year to $64 billion.This is roughly double the annual growth rate of the broader AI software market (31.4 percent CAGR), indicating that enterprise purchasing is shifting rapidly from standard application subscriptions toward underlying platform infrastructure.Second, sovereign artificial intelligence estimates exhibit wide dispersion due to differing boundary definitions. Narrower forecasts ($148 billion to $177 billion) isolate physical compute infrastructure and data localization hosting, whereas broader strategic evaluations, such as McKinsey's $600 billion estimate, encompass the complete software stack, custom model training, and specialized data curation services required for localized operation.Third, legal vertical technology spending serves as a key lead indicator for professional services adoption. The wide variance between conservative ($3.90 billion by 2030) and aggressive ($10.82 billion by 2030) legal tech forecasts highlights the ongoing strategic shift between standard SaaS tools and custom software development.Major enterprise commitments indicate that market expansion is being driven by custom internal builds rather than traditional off-the-shelf software licensing.Working Estimate Synthesis for Firm-Owned AIAssuming the United States maintains approximately 40 percent of global AI software spending, the domestic U.S. Applied-AI software market is projected to reach approximately $100 billion by 2027.Isolating the proprietary, knowledge-graph-driven segment from generic software consumption indicates that current U.S. corporate spending on firm-owned intelligence stacks sits in the low single-digit billions ($2 billion to $4 billion).However, as adoption expands past legal services into management consulting, corporate finance, specialized engineering, and healthcare, spending on firm-owned intelligence architectures is projected to scale toward $15 billion to $30 billion by 2030.This trajectory reflects a reallocation of enterprise capital away from commoditized multi-tenant AI applications toward sovereign cognitive infrastructure.Strategic Implications and Future OutlookThe expansion of the Proprietary Intelligence Foundry model alters enterprise IT management, capital allocation, and industry competition.Shifts in Capital Allocation and Enterprise IT ArchitectureFor decades, corporate IT strategies emphasized off-the-shelf SaaS applications to eliminate internal software maintenance costs and reduce custom code overhead.The Proprietary Intelligence Foundry reverses this trend for core knowledge functions.Because off-the-shelf models risk homogenizing business judgment, enterprise IT departments are resuming custom software engineering centered on ontology modeling, graph governance, and specialized AI model fine-tuning.This structural transition drives several key operational realignments:Enterprise capital is shifting away from generic per-seat software licensing toward sovereign cloud infrastructure, specialized data engineering, and knowledge graph construction.Simultaneously, execution relies on internal "Business Technologists" — multifaceted subject-matter experts embedded within operational units like marketing, finance, or legal — who direct the construction of domain ontologies that govern fine-tuning pipelines.Furthermore, enterprise software architectures are moving toward autonomous, agentic platforms that rely on structured knowledge graphs to navigate complex, multi-step business workflows without human intervention.Structural Realignment of Professional Services PricingIn knowledge-intensive industries, deploying firm-owned intelligence stacks challenges traditional billable-hour revenue models.As fine-tuned AI small language models execute complex analysis in seconds, firms bound to time-based billing will face fee compression if overall billable volume contracts.Conversely, organizations that successfully deploy Proprietary Intelligence Foundries can transition toward value-based or outcome-based fee structures.By utilizing proprietary AI engines to deliver expert-level legal, financial, or consulting work products instantaneously, these firms decouple revenue generation from human billable hours, driving higher profit margins while protecting their competitive moats.Synthesized Key ConclusionsThe strategic transition from generic public artificial intelligence tools to secure enterprise sovereign intelligence yields fundamental conclusions for strategic leaders:First, in an environment characterized by rapid AI model iteration, public foundation models represent ephemeral runtime layers. The primary durable asset is the governed, enterprise-owned knowledge graph that structures human tacit knowledge into a persistent digital capital asset.Second, over-reliance on public, vendor-hosted AI foundation models regresses enterprise outputs toward the statistical mean. The Proprietary Intelligence Foundry preserves pricing power and operational differentiation by training systems directly on private operational data and judgment heuristics.Third, employee knowledge harvesting systems must capture operational signals ambiently from daily work exhaust (negotiations, emails, document histories) to succeed in environments where billable hours or operational demands prevent manual data labeling.Reach out to learn more about our Applied-AI Initiative objectives.More...

In June, I wrote about why AI training had moved inside the enterprise: the open web has run dry of the cognitive density frontier models now need, and the next phase of the Applied-AI race would be won on access to human tacit knowledge, not public data.I said the leading tech vendors had already restructured to capture it. What I didn't have yet was a single, undeniable proof point of a professional services firm making the same bet with its own money.Now I do. Kirkland & Ellis, the world's highest-grossing law firm, is spending roughly $500 million of its own revenue over the next three to four years building proprietary AI systems it will own outright.A meaningful piece of that is a co-built platform with Palantir called the Fund Formation Engine, designed to run private-equity fund documentation, side-letter drafting, and obligation tracking across Kirkland's fund-formation practice.Kirkland didn't pick Palantir for its model. It picked Palantir for its ontology-modeling expertise — the ability to represent a firm's decisions, obligations, and relationships as a structured, queryable graph rather than a pile of indexed documents.That distinction is the whole story, and it's worth being precise about why.Text Retrieval Was Never Going to Be EnoughEvery enterprise AI deployment from 2023 to 2025 followed the same template: take a frontier model, bolt on a vector database, call it RAG.It works fine for "find me the clause that looks like this one." It falls apart on "what did we decide, and why, and what does that obligate us to now" — because a flat document index has no model of decisions at all. It just has text.The firms moving fastest past this limitation are building what I'd call a Proprietary Intelligence Foundry: a pipeline that captures the exhaust of daily work — email, meetings, negotiation threads, collaboration patterns — and turns it into a governed knowledge graph before any retrieval happens at all.That graph becomes the substrate for an AI training curriculum, which fine-tunes a smaller, specialized model that reasons over the firm's own judgment instead of averaging across everyone else's.The durable asset in this stack was never the model. It's the graph.Two Kinds of Collapse, and Why the Second One Should Worry You MoreAnyone paying attention to Applied-AI Initiative research has heard of model collapse — the well-documented risk that training generation after generation of models on synthetic data erodes the tails of the original distribution.It's real, but it's also manageable: recent work shows that accumulating synthetic data alongside real data, rather than replacing it, keeps the degradation bounded.The failure mode that should actually worry a firm like Kirkland is different. Call it knowledge collapse: the tendency of any shared, vendor-hosted AI model to regress everyone's output toward the statistical center of its training data.If every firm in a market routes its differentiated judgment through the same off-the-shelf AI assistant, the assistant quietly erases the differentiation. For a law firm, a consultancy, or a bank whose entire pricing power rests on judgment competitors don't have, that's an existential risk dressed up as a productivity tool. It's close to the exact language Kirkland used to explain why it built rather than bought: generic AI tools trained on broad market knowledge tend to converge on a lowest common denominator, and a firm charging $2,000 an hour cannot afford to sound like everyone else.The Constraint Nobody Wants to Talk AboutHere's the part that should temper any senior executive's enthusiasm before they green-light a similar build: none of this works if you assume your best people will happily spend un-billed hours training it.Roughly 90 percent of legal revenue still runs through the billable hour, a compensation model essentially unchanged since the 1950s. Knowledge-sharing isn't factored into partner comp at most firms, which means the people with the most valuable judgment to capture have the least incentive to sit still and hand it over. Client contracts are making this worse, not better — outside counsel guidelines increasingly bar firms from billing for AI-driven time savings outright, which means any ROI case for this technology has to be made through win rate, new revenue lines, or capacity redeployment. Never through hours saved.That's a genuinely awkward position for a firm to be in: forbidden from billing for the efficiency the tool creates, while still needing to justify a nine-figure investment in it.The firms getting this right aren't scheduling training sessions. They're mining signals their most skilled people already generate — email, meetings, document collaboration — without asking anyone to stop and teach the AI tool.It's the same passive-capture principle I flagged in June, now with a name and a price tag attached.The Takeaway for Every Other Knowledge-Intensive IndustryKirkland is a law firm, but nothing about this architecture is legal-specific.Any professional services business whose value is concentrated in a small number of senior people's judgment — consulting, investment banking, specialized engineering, high-end agency work — faces the identical structural choice: rent a generic AI assistant that averages your expertise into the market's, or build the unique graph that keeps it yours.The interesting white space right now is that almost nobody in Management Consulting has made this bet publicly yet. Every case study I could find sits in legal or corporate finance. That's either a sign consulting firms haven't caught up, or a sign the smart ones are building quietly and haven't said so publicaly.Either way, I'd bet on this thesis showing up in a consulting firm's own AI investment announcement within the next two to three quarters — and when it does, it won't be a coincidence that the forward-thinking firm making it competes on judgment, not headcount.For the AI architecture, vendor landscape and sourcing, see my Intelligence Foundry research and analysis.Reach out to learn more about our Applied-AI Initiative objectives.More...

Google just published the most detailed empirical picture yet of how Generative AI is actually being used across the global economy, and it should reset how executive teams talk about AI transformation.The AI & Economy ATLAS study, built on 15 million de-identified interactions across the Gemini App, Google AI Mode, and the Gemini API, maps usage against more than 800 occupations, 4,000 work tasks, and 150 countries.The headline finding is not that Applied-AI is transforming work. It is that AI has diffused everywhere while penetrating almost nowhere near as deeply as the enterprise boardroom narrative suggests.The AI Coverage IllusionGemini usage now touches 68 percent of detailed occupations, covering occupations that represent 88.4 percent of total U.S. employment.That is the statistic executives will quote. It is also the one most likely to mislead them. Among occupations with any measurable AI usage, the median worker is applying AI to just 21 percent of their constituent tasks.Only 3 percent of occupations show usage across more than 75 percent of their tasks. Breadth of adoption and depth of integration are two different metrics, and most enterprise AI strategy conflates them.Augmentation is Still Winning the ArgumentThe report's intent classifier found that end-to-end task automation accounts for less than 10 percent of AI conversations tied to non-routine cognitive work, the category that includes strategy, analysis, and creative problem-solving.Usage there clusters instead around drafting, review, and ideation.Routine cognitive work tells a different story: more than a quarter of those conversations target automation outright. The distinction matters for workforce planning. AI is not yet substituting for judgment.It is substituting for the codifiable parts of a job, which is a narrower and more predictable transition than most restructuring plans assume.The Wage Premium SignalThis is the finding I would put in front of any CHRO. A 1 percent increase in an occupation's median earnings correlates with a more than 2.5 percent increase in AI usage intensity, a relationship that survives controls for education.Weighted by conversation volume, the median salary among AI users runs close to $83,000, roughly $20,000 above the true employment-weighted national median.AI adoption is not spreading evenly across the workforce. It is concentrating among the few workers who are already the most valuable, which is precisely the talent stratification dynamic I flagged in the wake of recent business restructuring in the U.S. market.This latest research gives the skilled expert worker thesis a national dataset.The Off-Balance-Sheet Productivity GainOver 86 percent of all conversational AI usage happens outside formal work, and it correlates strongly with how people actually spend their non-work hours.The categories most over-represented relative to time spent are government services, legal topics, and financial administration, some by a factor of nearly twenty.Google's conservative estimate puts the annual unpaid productivity value of household AI use in the tens of billions of dollars in the U.S. alone, entirely invisible to GDP.For CFOs modeling AI's economic footprint, the workplace is only capturing part of the Applied-AI value creation story.Executive Outlook: The ROI PredictionNone of this supports the displacement narrative that dominates enterprise board-level anxiety, and it does not support the dismissive counter-narrative either.What it supports is a more disciplined read: AI is a general-purpose technology still in its shallow-adoption phase, distributing its early gains toward workers and geographies that already hold structural advantages. Adoption scales with GDP per capita almost one for one, and the lowest-usage quintile of countries, representing 17 percent of the global population, generates just 2 percent of AI conversations.If your enterprise Applied-AI roadmap assumes uniform diffusion across roles, geographies, or seniority levels, this data says otherwise. The organizations that win this decade will be the ones that treat task-level saturation, not headline adoption percentages, as the metric that actually predicts return on investment.Reach out to learn more about our Applied-AI Initiative objectives.More...

The Applied-AI Requirement, Seen from My Own DeskA few months ago, I began evaluating a paid license from some of the leading AI providers on the assumption that my advisory work would eventually outgrow the free AI app tiers.It has not happened yet. Each time I approached the point of subscribing, a new free or entry-level release arrived that adequately covered what I actually needed: drafting, research synthesis, and editorial refinement. My requirements were never exotic. They were representative of a large share of knowledge work, which is precisely the point. This is not a story about frugality. It is a story about an AI capability moving target.The capability that once justified a premium license a year ago is now embedded in the free tier of the same provider, or matched by a competitor's low-cost model.Independent benchmark trackers have shown the performance gap between open and proprietary models narrowing from double digits to less than one percentage point within a single year.For everyday business tasks, such as summarization, drafting, classification, and research synthesis, that residual gap is effectively invisible to the person doing the work. That's been my consistent experience.If a seasoned, one-person advisory practice cannot reliably justify a frontier license against a fast-improving free tier, the calculation is considerably harder for a large enterprise procuring subscriptions by the thousand.That is the requirement worth naming plainly: large enterprises are making multi-year AI spending commitments against a capability curve that resets every quarter. "Good enough" AI is always advancing.Practical Approaches Worth ConsideringThe instinct is to solve this with better AI model selection rules. Rules help, but rules alone will not survive a rapidly moving target. A few approaches deserve serious consideration.Tier by task, not by title. Access should follow the sensitivity, complexity, frequency, and business value of the work itself, not the seniority of the person requesting it.A default tier for routine work, a professional tier for power users, and a restricted tier for genuinely frontier-dependent tasks give finance and legal a defensible structure to stand behind. Essentially, a three-tier good, better, best approach.Assign ownership, deliberately. Line managers and their teams should not be expected to make model selection judgments on their own. Most were never given the cost visibility, the task-to-capability mapping, or the budget signal needed to do it well.A small central function, sometimes described in current research as an AI Orchestration role, should own the policy of matching models to use cases, while business units retain ownership of outcomes.Treat the mapping as a subscription, not a purchase. Because the underlying capability curve moves continuously, a model-to-task matrix written in January can be materially wrong by June.The organizations managing this well are reviewing their assignments on a recurring cadence, the same discipline applied to any other fast-depreciating vendor commitment.Design for portability before committing to an AI provider.An architecture built around a single named AI model creates the very lock-in that makes the moving target painful. An open architecture built around a capability requirement, with the provider and model as an interchangeable layer beneath it, absorbs a "the free tier just caught up" moment without a re-platforming project.The Pent-Up Demand for Just-in-Time MatchingThere is a reason Gateway and Model-Routing infrastructure has moved from a niche engineering conversation to a board-level budgeting concern in 2026.Enterprises are discovering that no static policy and no individual manager can keep pace with a model landscape that reshuffles its best-fit Applied-AI options every few weeks.What they are asking for, in effect, is a mechanism that performs the matching continuously and automatically: routing each request to the AI model that clears the quality threshold at the lowest defensible cost, then re-evaluating that routing as new releases arrive.The market signal here is not subtle. Cisco's own leadership has spoken publicly about the budgeting shock of treating the most powerful model as the default choice, and providers report usage surges on routing-oriented platforms that did not exist eighteen months ago.Tech industry analyst forecasts now put a majority of leading AI-driven enterprises on multi-model routing architectures within the next two years. That is not a forecast about technology adoption for its own sake. It is a forecast about enterprises trying to get ahead of an Applied-AI cost problem that has already arrived.For the C-suite, the runaway AI spend risk has two distinct sources, and it is worth naming both. The first is the familiar one: defaulting to the most expensive model out of convenience or unfamiliarity with the many viable alternatives.The second is less discussed and arguably more dangerous. As models get cheaper and faster, the friction that once limited how often employees queried them disappears, and usage volume can expand faster than the per-token savings, leaving total spend flat or higher even as unit costs fall.A just-in-time AI Gateway does not only route cost away from expensive models. It is also the only practical mechanism for keeping the resulting volume visible before it becomes a line item nobody can explain.None of this argues against adopting the very best models for your Applied-AI Initiative.It argues for treating "best frontier by default" the way a disciplined finance function already treats any other un-metered resource: valuable when the task warrants it, and a compounding liability when it does not.Reach out to learn more about our Applied-AI Initiative objectives.More...

The public internet, long treated as an inexhaustible resource for training large language models, has run dry. Not in terms of raw volume, but in terms of the cognitive density that frontier AI now requires.Research I published through GeoActive Group's Applied-AI Initiative confirms what a growing number of senior researchers have quietly acknowledged: the next phase of the AI race is being won or lost on access to human tacit knowledge, and the leading tech vendors have already restructured their organizations to capture it.This is not an incremental refinement to existing AI training methodology. It is a wholesale reorientation of how the most resource-intensive companies in the world are deploying their most valuable internal asset: the unwritten reasoning of their best people.The Structural Bottleneck Driving This ShiftThree converging constraints have forced this strategic pivot. First, models trained on generic web content have hit a reasoning ceiling. They perform adequately on surface-level tasks but struggle with deep domain execution, edge-case troubleshooting, and the multi-turn logical chains that autonomous enterprise agents will require.Second, the early reliance on low-skill, crowd-sourced annotation for Reinforcement Learning from Human Feedback (RLHF) has proven insufficient. Training an agent to write secure infrastructure code or diagnose complex system anomalies demands feedback from skilled practitioners at the highest level of their discipline, not general contractors.Third, the sovereign model ambition that now defines competitive positioning across the hyperscaler tier requires that models internalize not just what expert decisions look like, but why certain alternatives were rejected. That negative space, the path not taken, only exists inside expert human cognition.Key Applied-AI Research FindingsOur research identifies five distinct vendor implementation strategies, each revealing how differently structured organizations are solving the same underlying problem.Meta has undergone the most radical internal restructuring, transferring thousands of senior software engineers and product managers away from consumer products and into Applied-AI data generation roles. Its management architecture has been compressed to a 50-to-1 employee-to-manager ratio specifically to accelerate data iteration velocity. The internal logic is direct: elite corporate talent yields a higher intellect density for model training pipelines than any external contractor arrangement can replicate.Microsoft's approach centers on treating its own global workforce as a continuous behavioral data source. Its Customer Zero framework requires internal sales, HR, finance, and engineering teams to run pre-release AI agents in daily workflows, with every interaction logged as training data.Tools including Viva Skills and the Microsoft Graph are being used to map how expert employees handle context switching and solve layered business problems, capturing tacit metadata that no data lake or document repository could surface on its own.Google's strategy concentrates on its Site Reliability Engineering population, mining real-time debugging decisions, system architecture choices, and incident-response behaviors to create supervised fine-tuning data for specialized autonomous coding and infrastructure management agents.xAI draws its tacit knowledge across physical domains, pulling aerospace and autonomous driving edge-case telemetry from Tesla and SpaceX directly into Grok's training loops.OpenAI, lacking a comparable legacy engineering workforce, has built a hybrid approach around a global network of more than 100 deeply specialized external red team members combined with internal synthetic reinforcement learning teams that generate structured training data from existing frontier models.Enterprise C-Suite Executive OutlookThe strategic implications for enterprise leaders extend well beyond observing how hyperscalers train their models. Two pressure points deserve immediate senior executive attention.The first is intellectual property exposure. When senior engineers, underwriters, legal counsel, or financial analysts interact with third-party vendor AI tools in their daily workflows, the tacit problem-solving loops they generate may be contributing to a sovereign model that belongs to someone else.The value being extracted is not their data. It is their reasoning. Enterprises need governance frameworks that treat expert cognitive workflows as proprietary assets, not incidental byproducts of software usage.The second is the internal data strategy gap. Most organizations still define their AI readiness in terms of data lakes, document repositories, and transaction logs. That foundation is necessary but no longer sufficient.The competitive advantage in this next phase belongs to organizations that can capture the structured reasoning of their employee top performers within their own private infrastructure, and convert it into durable institutional intelligence -- rather than letting it dissipate when those individuals leave, retire, or are displaced by the very agents being trained on their knowledge.The tech vendors building Sovereign AI systems are not waiting for enterprises to understand this dynamic. They are already inside the workflow. The mandate for enterprise leadership is to decide, deliberately and soon, whether their organization's deepest expertise will remain their own.Reach out to learn more about our Applied-AI Initiative objectives.More...

Economic growth is fueled by strategic investment. Venture capital (VC) has quietly become one of the most consequential forces in the global networked economy, yet most C-suite leaders engage with it only at the margins.That needs to change.A landmark new report from the World Economic Forum and Stanford Graduate School of Business, released this month, offers a comprehensive and at times sobering assessment of where the VC industry stands and where it is headed.For senior executives navigating technology strategy, capital allocation, and competitive positioning, the findings carry direct implications.The Scale of the Opportunity and the Strain Beneath ItVC assets under management have grown more than sixfold since 2008, reaching $3.4 trillion globally in 2025. Seven of the ten largest companies in the world by market capitalization, including Apple, NVIDIA, and Amazon, received venture backing in their early stages.Among U.S. public companies founded in the past 50 years, VC-backed firms account for 94 percent of all research and development spending. These are not statistics about a niche asset class. They describe the financial architecture behind modern innovation itself.Yet the WEF report makes clear that the model is under significant structural pressure.Venture-backed companies are staying private for far longer than historical norms would suggest. The average time from founding to IPO now stands at 12 years, up from around 7 to 8 years in earlier decades. Approximately 1,920 unicorns, companies valued at $1 billion or more, remain privately held globally, representing over $7.3 trillion in post-money valuation and an estimated $3 trillion in unrealized value sitting on fund balance sheets.The downstream effect is a distribution drought.Among funds launched in 2021, three-quarters had returned less than 25 cents on the dollar by their fourth anniversary. Investors are, on average, receiving back only 60 percent of what they would typically expect at a comparable stage.Since 2022, U.S. venture funds have drawn a net $196.9 billion more from investors than they have returned, a capital deficit without precedent in the industry's history.Where Corporate Capital Enters the PictureThis is precisely where corporate involvement in venture has become more strategically relevant.Corporate venture capital (CVC) units have moved from peripheral experiments to core components of the innovation financing ecosystem. For large enterprises, CVC offers more than financial return.It provides early access to emerging technologies, direct sight lines into competitive disruption, and the ability to build ecosystem relationships before market dynamics force a reactive hand.The WEF report notes that the lines between traditional VC, growth equity, private equity, and corporate capital are increasingly blurring at the later stages of the funding lifecycle.Amazon, Alphabet, Meta, and Microsoft are collectively projected to spend more than $650 billion in capital expenditure in 2026, the vast majority directed at artificial intelligence (AI) infrastructure.These are no longer passive technology buyers; they are active co-architects of the venture ecosystem, funding the infrastructure on which the next generation of startups will be built.For mid-market enterprises and sector-specific corporations considering CVC programs, the message is clear: the window to establish credible, value-add relationships with emerging companies is narrowing as the largest players consolidate their positions.The Investment Trends Worth WatchingLooking ahead, three dynamics stand out as particularly consequential for business leaders.The rise of secondary markets represents the most immediate structural shift. Secondary transactions in U.S. venture reached $106 billion in 2025, nearly matching the total value of all VC-backed IPOs in the same year.While secondary activity remains concentrated around a handful of high-profile names, the underlying infrastructure is maturing rapidly. For corporate investors and family offices, this creates new entry points into private company stakes that were previously accessible only to specialist funds.Applied-AI Initiatives are rewriting the investment economics of the entire sector.In 2025, AI accounted for more than 50 percent of global venture deal value, with five companies alone absorbing 20 percent of all global VC capital. More critically, AI-native companies are reaching scale with a speed and capital efficiency that challenges conventional valuation frameworks.Some have crossed $100 million in annual recurring revenue in under a year. For corporations building technology strategy, this compression of development timelines changes both the threat horizon and the partnership opportunity.Geography will increasingly define competitive advantage.The WEF report reveals a striking disparity: in the U.S. market, unicorns reach $1 billion in valuation in an average of 3.4 years. Outside the U.S. realm, the median time is more than eight years.Regulatory harmonization initiatives in Europe and Southeast Asia are attempting to close this gap, but progress will be measured in years, not quarters.The institutions and enterprises that begin engaging substantively with these dynamics now, rather than waiting for the cycle to stabilize, will be positioned to shape the next era of innovation rather than simply react to it.Reach out to learn more about our Applied-AI Initiative objectives.More...

To date, the dominant narrative around artificial intelligence (AI) in business was one of cautious optimism shadowed by disappointment. Organizations launched pilots, generated buzz, and then quietly shelved initiatives that failed to scale.That narrative is changing. The World Economic Forum (WEF) inaugural MINDS report, produced in collaboration with Accenture, offers one of the most comprehensive snapshots yet of what successful, real-world Applied-AI adoption actually looks like.The findings are instructive, occasionally surprising, and carry clear strategic lessons for any organization still searching for the bridge between experimentation and ROI impact.The Scale of What is HappeningThe MINDS program drew applications from over 30 countries spanning every major region, with participation cutting across industries from energy and healthcare to financial services and advanced manufacturing.Information technology (IT) accounted for nearly one-third of all submissions, but what is striking is the breadth beyond that sector.Healthcare, automotive, retail, and battery manufacturing all featured prominently, signaling that AI has genuinely become a cross-functional, cross-industry force rather than a technology sector story.Perhaps the most telling demographic detail: more than 50 percent of applicants were small and mid-sized organizations with fewer than 500 employees. The persistent assumption that large-scale AI adoption requires large-scale budgets and headcounts is simply not borne out by the evidence.Innovation, it turns out, is not a function of size.The Numbers That Demand AttentionThe report's impact tables are where the real story lives. Several figures stand out.CATL, the Chinese battery manufacturer, used a physics-informed AI platform to compress design timelines from two weeks to minutes, cut prototype development cycles from 24 to 13 months, and achieve annual R&D savings of $140.6 million.Meanwhile, Fujitsu's real-time supply chain optimization system delivered $15 million in reduced annual inventory costs and $20 million in stock reduction for a single client deployment.In healthcare, Ant Group's multi-modal clinical AI platform achieved over 90 percent diagnostic accuracy across more than 5,000 disease categories while serving 160 million users and nearly one million doctors.Landing Med's AI-assisted cervical cancer screening now covers 91 percent of China's remote provinces, in a country where fewer than ten pathologists per million people are available for this task.The Ministry of Health of Saudi Arabia, partnering with Amplifai Health, achieved a twelvefold increase in screening capacity for diabetic foot conditions while reducing treatment costs for patients by 80 percent.On the infrastructure side, State Grid Corporation of China's city-scale AI platform for Shanghai's power grid generated over $1.12 billion in avoided construction costs and eliminated 510,000 tonnes of carbon emissions annually.They're not theoretical projections. They're documented outcomes from deployed systems.Five Insights Separating Leaders from LaggardsThe WEF report distills its findings into five interconnected insights, and together they form a coherent theory of successful AI adoption. The most advanced organizations are treating AI not as a tool to be plugged into existing workflows, but as a strategic capability that reshapes how they compete.Roughly 75 percent of MINDS applicants reinvest returns from current AI projects to fund new adoption, rather than treating early wins as endpoints.Human capital emerges as the decisive variable. Organizations that co-designed AI solutions with frontline employees, and invested in role-based upskilling alongside deployment, consistently outperformed those that treated AI as a purely technical implementation.Foxconn's Project Genesis is an instructive case: decades of manufacturing expertise from experienced workers were systematically digitalized and paired with AI agents, producing a 50 percent reduction in changeover workload and a 30 percent decrease in problem resolution time.Data quality remains the most commonly cited barrier, but the report reveals that extensive datasets are not always a prerequisite. UCSF and SandboxAQ used physics-based simulations to generate high-quality training signals, enabling screening of 5.6 million drug compounds in weeks rather than years, a 36-fold reduction in experimental effort.Where Applied-AI Growth Will ConcentrateThree trends from the WEF report point toward near-term growth opportunities.First, agentic AI is moving from novelty to norm.More than one-third of MINDS applicants were already deploying agentic systems, and the use cases, from hospital bed management to semiconductor chip design, suggest this technology is ready for high-stakes enterprise environments far sooner than many anticipated.Second, edge computing is becoming a strategic differentiator.Hyundai and DEEPX demonstrated that custom AI silicon can deliver GPU-level inference at 70 percent lower power consumption, opening entirely new categories of autonomous, on-device applications in robotics, logistics, and smart manufacturing.Third, organizations that treat responsible Applied-AI as a governance checkbox rather than a design principle will increasingly find themselves at a competitive disadvantage.The emergence of what the report calls "trust-by-design" architectures, where explainability, bias detection, and compliance are embedded directly into AI systems rather than layered on afterward, is quietly becoming a baseline expectation in regulated industries.Next Steps: Apply the Lessons LearnedIn Summary, the forward-thinking organizations profiled in the MINDS cohort have moved the conversation decisively toward strategic business outcomes.The question for everyone else is no longer whether AI can deliver measurable value. It clearly can. The better question is what combination of strategy, culture, data discipline, and AI infrastructure investment will determine who captures that value, and how quickly.Reach out to learn more about our Applied-AI Initiative objectives.More...

The generative AI (GenAI) wave that began with ChatGPT's arrival in late 2022 has already started to feel like yesterday's story.A recent TBR research report on the Applied-AI and GenAI market landscape makes one thing clear: the industry is pivoting fast, and the companies that fail to adapt to agentic AI will find themselves playing catch-up in a market that rewards those who move decisively.For the uninitiated, agentic AI refers to systems that don't just respond to prompts but actively plan, execute, and iterate across complex multi-step workflows with minimal human intervention.This is no longer a futurist talking point. It is reshaping how enterprises think about automation, how IT service firms price their work, and how hyperscalers compete for the next trillion dollars in technology spending.A Market Growing at Breakneck SpeedThe numbers alone make a compelling case for attention.TBR estimates that combined AI and GenAI revenue across major hyperscalers, including AWS, Microsoft, Google, and Oracle, reached $46 billion in 2025, representing a year-over-year increase of 73 percent.Capital expenditure projections show that figure climbing steeply toward 2027, with infrastructure investment accelerating in parallel. These are not incremental gains. They signal a fundamental rewiring of the business technology transformation economy.A significant portion of this revenue surge is being driven by AI model developers themselves. Companies like OpenAI and Anthropic are securing enormous infrastructure commitments from cloud providers to support both current training workloads and anticipated future demand.TBR flags this as a concentration risk worth monitoring closely.Should sentiment shift or financing conditions tighten, the gap between backlog projections and recognized revenue could widen considerably. For now, the momentum holds, with over 77 percent of enterprise respondents in TBR's cloud customer survey reporting that AI had exceeded their value expectations.Commercial Alliances Are Being RedrawnBeyond the headline Applied-AI revenue figures, one of the most strategically important developments is the restructuring of alliance ecosystems.Large IT services firms are no longer operating from a technology-agnostic partner posture. They are making deliberate, named bets on best-of-breed AI vendor relationships.HCLTech, for example, has moved to co-develop and co-sell AI-enabled industry solutions with hyperscalers, Databricks, and Snowflake, launching at least eight such solutions including an AWS-based financial services tool called InsightGen.Meanwhile, Kyndryl has shifted from bilateral partnerships toward orchestrated, multiparty alliances, combining AI capabilities with infrastructure expertise through deals involving HPE and NVIDIA.Microsoft, for its part, has broadened beyond its OpenAI relationship toward a multi-model strategy that now includes Anthropic's Claude models on Azure alongside tighter governance frameworks built with Workday.The pattern here is consistent: depth over breadth, co-creation over reselling, and governance as a competitive differentiator rather than an afterthought.The OEM and Telecom OpportunityTwo segments that deserve closer attention from investors and strategists are original equipment manufacturers (OEMs) and communications service providers (CSPs).On the OEM side, on-premises and hybrid AI deployments are entering a slow but steady ramp-up phase. Enterprise customers pursuing these configurations tend to require more comprehensive services engagements, covering AI advisory, lifecycle management, and industry-specific deployment support, often built around NVIDIA AI Enterprise frameworks. The constraint is that service provider customers still dominate OEM AI server revenues in 2025, limiting the addressable professional services market for now.For telecoms, TBR's projections are striking. The total potential annual AI-related value to CSPs could reach $170 billion by 2030, split roughly between $90 billion in new revenue opportunities and $80 billion in cost efficiencies.Early evidence of new revenue materializing is visible in network transport deals won by Lumen and Zayo and in exploratory infrastructure co-location efforts by Verizon and AT&T.AI Investment Strategy: The Road AheadThe TBR framework for agentic AI evolution across a three-year horizon is instructive for anyone planning technology investment strategy.Today, agents handle simple, low-variable tasks but falter on complexity, with memory that rarely persists beyond a single session. Within one to two years, multi-hour and multi-day workflows become viable, governance layers standardize, and inference costs fall.By 2028 and beyond, the vision is one of domain-specialized agents acting as persistent digital workers, coordinating in teams, and managing end-to-end processes with only periodic human oversight.The organizations best positioned to capture this value are those investing now in orchestration infrastructure, evaluation tooling, and the services capability to manage not just individual agents but entire agent populations.The enterprise winners will not simply be those who adopted AI earliest. They will be those who built the operational discipline to scale it responsibly and profitably.The agentic era is not coming soon, it has already begun.A 2026 Agenda for the Enterprise C-SuiteFor large enterprise leaders, the remainder of 2026 is not a period for continued experimentation. It is a period for commitment.The window to establish durable AI operating models before competitors lock in structural advantages is narrowing, and several priorities demand executive attention now.The first is governance. As agentic systems move beyond isolated pilots into operational workflows touching finance, HR, supply chain, and customer engagement, the absence of clear accountability structures becomes a serious liability.CEOs and boards must demand that CIOs and CTOs present coherent governance frameworks covering how agents are evaluated, audited, and corrected when they err. This is not a compliance checkbox. It is a foundation for scaling with confidence.The second is vendor strategy. The alliance restructuring underway among IT services firms and hyperscalers is not background noise. It reflects a market in which multimodel, multiparty ecosystems are becoming the standard architecture for enterprise AI delivery.C-suite leaders should be asking whether their current vendor relationships give them flexibility across model providers, or whether they are locked into a single stack at precisely the moment when the competitive landscape is diversifying.Renegotiating or broadening those agreements in 2026, while leverage remains available, is preferable to doing so under pressure in 2027.The third is talent and services sourcing. The TBR data makes clear that most enterprises will not be able to deploy advanced AI solutions in-house without significant external support, particularly in the on-premises and hybrid deployment scenarios growing in strategic importance.Building relationships with services partners who have reusable agent frameworks and domain-specific accelerators, rather than those offering bespoke implementations alone, will determine how quickly and economically an organization can move from pilot to production.Finally, CFOs in particular must address the ROI measurement gap. The 77 percent of enterprises reporting that AI exceeded their value expectations is an encouraging signal, but optimism is not a budget justification.Establishing clear metrics for agent performance, cost-per-action baselines, and productivity benchmarks before the next budget cycle will separate organizations that can defend and grow their AI investments from those that find themselves retreating under pressure from skeptical boards.The agentic era rewards those who plan deliberately, and moves with conviction.Reach out to learn more about our Applied-AI Initiative objectives.More...

We have now entered the AI Agentic era, according to the latest series of reports by Google's artificial intelligence (AI) researchers.The shift from passive generative AI models to autonomous AI agents that can plan, reason, and act on our behalf is the most profound digital transformation in decades.As Applied-AI Initiatives replace deterministic code, a significant challenge has emerged.Building an AI agent is easy; however, trusting it is complex.The current AI market momentum reveals a stark last-mile gap.While a developer can spin up an AI prototype in minutes, roughly 80 percent of the effort required to reach production is consumed by the work of safety, validation, and infrastructure.The reason is simple: AI agents are non-deterministic. They can pass 100 unit tests but fail catastrophically in the field because of a flaw in their judgment, not a bug in the code.Core Architecture and the Problem-Solving LoopAn Applied-AI agent is defined by the synergy of four components:The Model (reasoning brain), Tools (actionable hands), the Orchestration Layer (governing nervous system), and Deployment (the physical infrastructure).The 5-Step Loop: Agents solve problems by cycling through getting a mission, scanning the scene for context, thinking through a plan, taking action via tools, and observing results to iterate.Taxonomy of Autonomy: Agentic systems scale from Level 0 (isolated reasoning) to Level 4 (self-evolving systems capable of creating their own tools and sub-agents).To bridge the trust gap, we must embrace three primary insights from Applied AI results.First, the Trajectory is the TruthIn the world of AI agents, the final answer is merely the last sentence of a long story. To judge an agent's quality, we can no longer just look at the output (the "Black Box" result).We must inspect the reasoning trajectory — the "Glass Box" view of the AI agent’s internal monologue, its tool calls, and its reaction to environment changes.For example, if an AI agent takes twenty steps to book a flight when it should have taken three, it is a low-quality agent, even if it eventually succeeds.Second, Context is the New CodeBecause agents are stateless, their "intelligence" is entirely dependent on the information we pack into their context window — a process now called Context Engineering. We must distinguish between the AI "research librarian" (RAG), which provides global facts, and the "personal assistant" (Memory), which tracks user-specific nuances.A truly intelligent AI agent doesn't just know the world; it learns and adapts to you over time.Third, We Must Transition to AgentOpsTo manage an autonomous AI fleet, organizations need a continuous, self-reinforcing loop: the Agent Quality Flywheel.This means instrumenting every agent from the first line of code to emit the logs and traces needed for judgment.Every production failure must be captured and programmatically converted into a new test case for a Golden Evaluation Set.This ensures the AI system doesn't just run; it evolves.Finally, we must acknowledge that the human is the ultimate arbiter.Automation, from LLM-as-a-Judge to safety filters, provides scale, but the definition of "good" must remain anchored in human expertise and values.AI can grade the test, but humans must write the essential rubric.In summary, Google's researchers found the organizations that win this era will be those that move beyond the hype of clever demos and invest in the rigorous architecture of trust.The future is agentic, but its success will be determined by our ability to see inside the AI agent's mind and ensure it remains a reliable, safe, and efficient business partner.Reach out to learn more about our Applied-AI Initiative objectives.More...