The KPI-backed read on Agentic AI Enterprise — Adoption, Readiness & ROI.
The State of Gen AI Leadership 2025-26 (Volume I) is Product-Led Hub’s KPI-backed read on enterprise AI — 175 metrics from 17 sessions and 50 leaders at the Gen AI Leadership Forum in Athens, as seen on our stage. The KPIs in this first volume are drawn from 2025 for now; later volumes carry the index forward. It maps the gap between enterprise AI adoption and real enterprise AI readiness — only 1 in 5 organisations are genuinely AI-ready — and the shift toward the State of Agentic AI, cross-checked against independent global research.
// The only KPI-backed view of how enterprises are moving from AI adoption to real AI readiness — and toward the agentic frontier.
Two-thirds of companies’ AI effort goes into just two things — putting AI to work in real use-cases, and proving it pays off. Almost nothing goes into the harder groundwork: strategy, wider adoption, and next-generation ‘agentic’ AI. It shows in the results — where AI is actually deployed it delivers, yet only about 1 in 5 organisations has the data, skills and governance to scale it past a pilot. The effort is racing ahead of the foundations — and that gap decides who turns early wins into lasting change. Throughout, that readiness is scored 0–5 across eight pillars — the AI Delta Maturity Framework™ — so ‘ready’ always means the same thing.
Five years ago, “AI strategy” was a slide in a board deck. Today it’s the operating system of the business — or it should be.
We watched that shift happen in real time. In November 2023, we opened the first AI Summit at the Athens Conservatoire — a room testing whether Greek business leaders wanted a serious conversation about applied AI. They did.
Five years and 12 summits later, the Disrupt (AI) Summit Series has put 200+ speakers in front of 4,000+ attendees across 700+ companies — from that first cautious gathering to a full calendar spanning Athens, Ioannina, and every function from engineering to the boardroom. We’ve watched the conversation move from “what is Gen AI” to agentic systems that plan, act and close the loop on their own.
That’s five years of signal — the kind that usually lives and dies in a conference room, in a slide someone photographed and never opened again.
This year we decided not to let it disappear. We’ve sat through enough AI pitches to know how much of this space runs on opinion — so we didn’t ask the room what it believed. We logged what it measured.
Every figure a leader stated on stage or showed on a slide at the Gen AI Leadership Forum 2025 edition, captured and cross-checked against IBM, McKinsey, Gartner, WEF and Eurostat. Not our take. Not a vendor’s roadmap. Fifty leaders, on our stage, and the numbers they actually stood behind.
This is Volume I of this report. It’s not the final word — it’s the first entry in an index we’ll keep building, Forum by Forum, to track how organisations actually move from adopting AI to running it autonomously.
Read it before your next board meeting. Then come close the loop live at the 2nd Gen AI Leadership Forum, October 21–22, at OTEAcademy, Athens.
What this report is. The State of Gen AI Leadership 2025-26 (Volume I) is original research by Product-Led Hub, drawn from the Gen AI Leadership Forum 2025 — 175 KPI metrics leaders stated on stage or showed on slides across 17 sessions, turned into an evidence-based view of how enterprises move from AI adoption to real AI readiness.
Part of a bigger picture. This is Volume I — its KPIs and results reflect 2025 for now — one edition in an ongoing series across Leadership, Engineering and Product — that compile into The State of Agentic AI (2023–2026), Product-Led Hub’s evolving, KPI-backed index of how organisations move from adoption to execution. Each edition adds a lens; together they build the full picture.
Why it was crafted. Most talk about “AI leadership” runs on opinion, vendor decks or borrowed forecasts. We built this to replace that with evidence — a hard baseline of what leaders in the room actually measure, so any organisation can benchmark itself honestly and see the real distance between adoption and execution.
How it was built. Every figure was coded by hand from the sessions and cross-checked against independent global research — not surveyed opinion or resold analyst forecasts. (Full method in the Methodology section.)
Product-Led Hub (PLH) is a product and AI education and events company based in Athens, Greece. Founded in 2019, PLH designs and delivers certification programs and corporate AI upskilling engagements for professionals and organizations, produces industry summits across Greece, and cultivates a professional community connecting product and AI practitioners, leaders, and organizations across the region.
This report is drawn from one of those events — the Gen AI Leadership Forum 2025. PLH doesn't survey opinion or resell analyst forecasts; it works from primary, practitioner-led signal. Across a series that now runs to 12 summits · 150+ talks · 200+ speakers · 4,000+ attendees, every figure is coded by hand. The same teams PLH trains and convenes are the ones whose work this report measures — which is why PLH can read the move from prompting to autonomy from the inside.
The stage is reserved for people who have built and run real AI systems in production — not analysts or vendors pitching a roadmap.
PLH measures what reached production: maturity, governance, cost and failure modes are reported, not hidden behind a highlight reel.
Since 2023, the Disrupt (AI) Summit Series has grown into a series of events addressing the challenges and opportunities that senior AI and technology leaders encounter across diverse business sectors. The 2026–2027 season marks its 5th anniversary — a milestone reflecting five years of bringing together global technology leaders. Beyond exploring what Gen AI makes possible across fields, each event delivers a comprehensive view of AI's role in tech and product teams — culminating in a focus on how leadership itself is reshaped by these advancements.
Featuring in-person technical workshops, engaging sessions, interactive roundtables, live product demonstrations, detailed case studies, and discussions with industry peers and international thought leaders, the Disrupt Summit Series is the premier destination for AI innovation.
Where leadership meets the intelligence economy. We are entering a new paradigm where competitive advantage is no longer defined by access to AI — but by the ability to execute with it at scale. This is the shift from adoption to execution. From isolated use cases to system-wide intelligence. From pilots to production. From potential to measurable impact.
AI is no longer a distant frontier — it is the operational core of modern organisations of every scale. Today’s challenge is no longer understanding AI; it’s embedding it into the fabric of decision-making, systems, and execution. The Gen AI Leadership Forum sits exactly at this inflection point — where executive leadership, founders, operators, and investors meet the operational reality of GenAI-driven transformation.
AI leadership will not be defined by technology alone, but by how organisations of every size redesign themselves around it — structurally, operationally, and strategically.
Give it a prompt, it returns an output — text, an analysis, code. Powerful, but it waits for instructions and stops at the answer.
Give it a goal. It plans, uses tools, takes actions and adjusts — looping until the job is done. Value moves from a good answer to a completed task.
The Forum is a single Athens event with a largely Greek enterprise audience, but every headline is cross-checked against global research (IBM, McKinsey, WEF, Eurostat, Gartner).
All 175 KPI metrics are extracted verbatim from timestamped session transcripts; 10 were cross-checked against named external sources.
Some figures are speaker-cited, anecdotal or illustrative — preserved with their original framing.
No. The coded data ends in 2025. Forward figures are external forecasts (Gartner, WEF) placed next to the Forum's trajectory — directional, not PLH measurements.
The readiness gap: 60–70% of organisations run AI in at least one unit, but only 1 in 5 are genuinely AI-ready to scale.
Around 60–70% of organisations now run AI in at least one business unit — but broad, enterprise-wide AI adoption is far rarer.
Only about 1 in 5 are genuinely ready to scale AI — the enterprise AI readiness gap between running AI and being ready to run it at scale.
Where it’s applied well, early adopters report meaningful returns — up to 21% better decisions, ~15% higher productivity and roughly 2 hours freed per employee.
A unified profile of every attendee across all 17 sessions — seniority, function, company size and industry.
Three in four people in the room (74.9%) were C-level or senior management — this was an audience of decision-makers, not practitioners.
Engineering was the single largest function (38.8%), but a broad non-technical tail — leadership, product, data, and a mix of HR, marketing, sales and operations — shows AI being treated as a cross-functional agenda, not just an IT project.
The overwhelming majority came from mid-sized and large organisations, where the gap between adopting AI and being ready to scale it carries the most operational weight.
Software and IT led at 38.8%, but finance, industry/energy and a long tail of other sectors made this a cross-industry conversation rather than a tech-only one.
The 50 leaders who took the stage across two days.
~28% work for a company headquartered outside Greece, the US being the dominant external centre (best-effort classification by primary HQ).
The line-up skewed senior and hands-on: founders, C-level leaders and engineering heads set the tone, with more than a quarter representing companies headquartered outside Greece.
George NikolaropoulosCTO, Natech Banking Solutions
Ioannis PapikasCTO, E-Satisfaction | Lead Product Trainer, Product-Led Hub
Despina KitsouAI Product Manager (AIOps), Nokia
Maggie AthanassiadiDirector of Technology, Innovation & Digital Transformation, SEV Hellenic Federation of Enterprises
Aleksandar PreradovicManaging Director and Chairman, Greece, Cyprus & Malta, Dell Technologies
Yannis RizopoulosJournalist
John SalichosCTO, ENTERSOFTONE
Sotiris KaragiannisChief Innovation Officer, Space Hellas
Marilena NikouHead of Product Marketing, ClearSkies
Thanasis PolitisSolutions Architect, Public Cloud Group - PCG
George MarinosAssistant General Manager Innovation & Digital Partnerships, National Bank of Greece
Ioannis TsilirasData & AI Transformation Director, Hellenic Telecommunications Organization (HTO)
Antonis ApergisChief Information Officer, Generali Hellas
Evangelos VagiasTechnology Director, DEMO Pharmaceuticals
Nick KontonicolasGroup Digital Technology & Operations Director
John MichailidisHead of Big Data & Engineering, Synthetica PC
Simon SeeGlobal Head Nvidia AI Technology Centre, NVIDIA
Panos BassiosDigital, Data & Analytics Director, MSD
Tilemachos MoraitisHead of Government & Corporate Affairs, Microsoft Greece
Vasileios RovilosHead of EU Policy, Credo AI
Aliki FoinikopoulouSenior Director, Global Public Policy, Salesforce
Vassilis KoutsoumpasDigital Policy & AI Adviser to the Prime Minister, Government of the Hellenic Republic
Thanasis NavrozoglouCEO, Natech Banking Solutions | Co-founder & Vice Chairman, Snappi
Emmanouil SerrelisGroup CTO, Alphabet Education Group
Kostas KetsietzisBusiness Editor, Insider
Dimitris TsingosCo-Founder, President & CEO, Epignosis Learning Technologies
Stelios LelisChief Data & Risk Officer, Optasia
Dimitris PapastergiouMinister, Ministry of Digital Governance of Greece
Nicholas RodopoulosSpecial Secretary SEPE / President & CEO, OnLine Data SA
Vassilis KarkatzounisSpecial Secretary for Artificial Intelligence and Data Governance, Hellenic Ministry of Digital Governance
Prof. Dr. Lilian MitrouProfessor, University of the Aegean / ILT-EPLO
Christos ZanganasManaging Partner, Skopa - Zanganas & Associates Law Firm
Christos LeivadiotisEnterprise Architect for Cloud and AI, Coca-Cola Hellenic
Marlen KarakoliHead of Modern Work Solutions, Office Line SA
Stavros MenegosChief Product Development Officer, ENTERSOFTONE
Manolis KatsifarakisDirector of AI, Epignosis Learning Technologies
Mihalis RikakisCEO, Dikaio.ai
Vassilis GalakosCEO, Comsys
Dimitris TogiasCTO, Instacar
Gerasimos MichalitsisCEO, Scrufy Robotics
Eleftherios AntoniadesFounder & CTO, Odyssey Cybersecurity & ClearSkies
Myrto PapathanouPartner, Metavallon VC
Alex EleftheriadisPartner, Big Pi Ventures
Tasos VasiliadisFounder, Joist
Nikos AntoniouFounding Partner, Apeiron Ventures
Yannis ImelosCo-Founder & CEO, Linq
Alexandros BasakidisCo-Founder, Agile Actors
Alexandra LoiChief People & Sustainability Officer, ESL FACEIT Group - EFG
Nikos BozikisHead of Global Talent Acquisition, Metlen Energy & Metals
Nikoletta MerziotiChief Operating Officer, Sleed
The reads that hold across all 175 KPI metrics. Each finding sets what our AI Delta Maturity analysis measures against what the wider industry shows — then states plainly what the two establish together.
The verdict
Enterprise AI readiness stands at just 1.61 out of 5.
Across eight readiness pillars, the field is barely a third of the way to ready — an AI Delta of 3.4 still to close.
Global AI investment will exceed $300–400B in 2025, up from $235B and growing 10× faster than the economy.
The money is racing far ahead of the readiness. Capital is being committed faster than organisations can absorb it — the spend is real; the capacity to convert it into scaled value is not.
External check — agreesMIT finds 95% of enterprise GenAI investment yields no measurable return, and Deloitte flags the same rising-spend, elusive-returns paradox. MIT / Forbes → · Deloitte →The foundation
Three readiness pillars score an outright zero.
Strategy & Leadership, Data & Infrastructure and Innovation Velocity carry no measured evidence of readiness whatsoever.
Microsoft, Google and AWS are ordering hundreds of thousands of next-generation GPUs.
The industry is stockpiling compute at scale — yet the strategy, data foundations and forward-capability to use it do not exist. This is hardware with no ground to run on.
External check — agreesMcKinsey names data quality the #1 barrier to scaling AI; Gartner warns most organisations lack AI-ready data. McKinsey → · Gartner →The imbalance
Execution has outrun the foundation beneath it.
The single strong pillar — Use Case Deployment at 4.17/5 — rests on a data-and-infrastructure base scoring 0.00.
The cost of inaction is severe: offshore downtime runs $500K an hour, equipment failure $200–500K a day.
The pressure to deploy is undeniable, so organisations push AI live regardless — on a base that cannot hold it. Deployment sprints ahead; the foundation is left behind.
External check — agreesMcKinsey: adoption is near-universal but only about a third have scaled; MIT: 95% of pilots stall on a weak base. McKinsey → · MIT / Forbes →The bright spot
Where AI is deployed, it delivers — the lone proven strength.
Use Case Deployment (4.17) dwarfs every other pillar (next best 2.50): 50+ live applications, −40% downtime, −30% scrap, +20% customer satisfaction.
The prize is quantified — a quarter of working hours are lost to repetitive tasks; 15% of legal spend is avoidable.
The returns are proven and measured, not speculative. The value is on the table for any organisation able to deploy and scale it properly.
External check — agreesMcKinsey reports 10–20% function-level cost cuts where AI is deployed; Deloitte finds 74% of top initiatives meet or beat ROI (20% above 30%). McKinsey → · Deloitte →Partial progress
The remaining pillars are partial — moving, not mature.
Adoption, Business Impact and Governance sit at 2.0–2.5, with Talent just below at 1.67 — genuine progress in each, held back by equally real gaps.
Training is scaling — but real in-house capacity is concentrated in a few enterprise giants, not the market.
Activity keeps rising, but volume is not becoming capability. These pillars are heading the right way — they have not matured.
External check — agreesWEF: training completion is rising (41%→50%) but a 39% skills gap persists; Korn Ferry projects 85M roles unfilled for lack of skills. WEF → · Korn Ferry →The gap
Adoption has decisively outrun readiness.
60–70% of organisations run AI live in at least one unit; only 1 in 5 are genuinely ready to scale — a 40-to-50-point gap, and the dataset’s most consistent structural finding.
AI keeps getting cheaper — inference costs half of last year’s, entry pricing near $20 a month.
Falling costs make switching AI on easy and cheap. Being ready to scale it is a harder, different thing — and four in five organisations are not there.
External check — agreesMcKinsey: ~88% use AI, only ~6% see enterprise-wide impact; MIT: 95% of pilots show no P&L impact. McKinsey → · MIT / Forbes →Attention is not readiness
The loudest signals are not the weakest pillars.
The most-measured theme and the noisiest topics are not where readiness is lowest. Attention and readiness are not the same measure.
Capital is everywhere — VC funds multiplying, the EU pledging €200B for AI.
The volume of money and coverage is easily mistaken for progress. It is not readiness — and it usually points away from where the real gaps sit.
External check — agreesMIT’s ‘GenAI Divide’ — high adoption, low transformation; McKinsey: widespread use, rare real impact. Noise isn’t readiness. MIT / Forbes → · McKinsey →The loud worry
Governance is the loudest worry — but mid-tier, not last.
It scores 2.50/5. The true zeros are strategy, data and forward-capability — not governance.
The threat is real: the average breach costs $4.44M and cyberattacks rose 44% in a single year.
Governance dominates the room because the threat behind it is real and expensive. But loudness is anxiety, not the lowest score — the quiet gaps are the deeper ones.
External check — agreesIBM: 63% of organisations have no AI-governance policy; Gartner expects 40% of CIOs to demand ‘guardian agents’ by 2028. IBM → · Gartner →A caveat
Every win is a single, well-scoped deployment.
The strong results are real but pilot-level — none is yet proven enterprise-wide.
Many of the boldest figures are explicitly illustrative — benchmarks stated as goals, not results achieved.
The wins are genuine and repeatable at pilot scale. The leap to enterprise-wide, proven-at-scale results has not been made.
External check — agreesMIT: only 5% of pilots reach real scale; McKinsey: only about a third have scaled AI enterprise-wide. MIT / Forbes → · McKinsey →This report is original research by Product-Led Hub — a KPI-backed evidence base on the state of AI leadership across the enterprise. This is Volume I; the figures here are drawn from 2025. Every number in it is a specific figure a leader stated on stage or showed on a slide during the Forum, captured and checked by our research team, never estimated or averaged. In total we recorded 175 such figures across 17 sessions over two days. Here is how we built and checked that dataset, so you can see exactly how much weight each figure carries.
We didn't stop at recording what was said. Key figures were cross-checked against published research from named outside organisations — IBM, McKinsey, Eurostat, the World Economic Forum and Gartner among them — looked up live as this report was written, to confirm the Forum's numbers held up against global benchmarks.
The report reads the same 175 figures two ways, and it helps to keep them apart:
Each pillar's score is the share of its readiness KPIs that are positive: score = ready ÷ (ready + gap) × 5. The distance to a full 5 is the pillar's AI Delta (Δ). Only KPIs that measure the field itself are scored — of the 175 KPIs, 22% are readiness-relevant (ahead or short), which dedupe to the distinct scored KPIs the pillars are scored on; the other 78% are vendor product results or market/investment backdrop, shown alongside but never scored. A pillar is flagged thin when it rests on no corroborated Ready or Gap evidence — the same rule the AI Delta Maturity Framework™ applies, and why the three zero-scoring pillars read as zero.
We didn’t invent these eight readiness pillars. They adapt the AI-maturity models the industry already uses — Gartner, McKinsey, Sema4.ai, Straive and Nemko — organised into eight capabilities after the pattern of Nemko’s AI-CMM: strategy · talent · data & infrastructure · deployment · adoption · business impact · governance · innovation velocity. Two of the eight — Adoption & Usage Depth and Innovation Velocity — are PLH’s own synthesis and the thinnest-sourced externally, as the framework notes.
The State of Gen AI Leadership 2025-26 (Volume I) is original research by Product-Led Hub, researched, hand-coded and written by Despina Exadaktylou — Founder & Managing Director, Product-Led Hub. Every KPI was captured from the sessions of the Gen AI Leadership Forum 2025 and cross-checked against named external sources (IBM, McKinsey, WEF, Eurostat, Gartner).
You’re welcome to reference, quote and share these findings — with attribution to Product-Led Hub and a link back to this report. Please don’t republish the full report, datasets or charts without written permission.
Product-Led Hub (2025). The State of Gen AI Leadership 2025-26 (Volume I). productledhub.com/state-of-gen-ai-leadership-2026
The 175 KPI metrics, first counted two ways — across the six agenda themes and the fourteen KPI types — then each opened up into what was actually measured.
Two themes — integrating AI and chasing ROI — draw more than half the numbers, while strategy and the agentic frontier barely register. And one measurement type, Transformation, is a quarter of everything measured.
Every KPI is also read as one of four kinds of evidence — the cut that decides what actually feeds the readiness score.
The most important cut: only 22% of the dataset (Ready + Gap) actually measures whether the field is ready. The other 78% is what suppliers can do and what the market is spending — kept as backdrop, never scored. This is why a loud topic can still hide a thin foundation.
Each theme (classified by topic) opened up by the KPI types inside it.
Each type's metrics grouped by what they actually measure — read from the descriptions. The two broadest are charted; the rest are condensed.
The AI Delta (Δ) is the measurable gap between where an organisation’s AI capability is today and where it needs to be to scale.
The AI Delta Maturity FrameworkTM is Product-Led Hub’s original model for measuring how ready an organisation is to scale AI. It scores eight readiness pillars — strategy, talent, data & infrastructure, use-case deployment, adoption, business impact, governance, and innovation velocity — each 0 to 5, on the weight of evidence that the capability is genuinely in place. The distance from a pillar’s score to a full 5 is its AI Delta (Δ) ; averaged across the eight, it gives a single readiness score and an overall Delta. Any organisation can run it on its own evidence. It synthesises and extends established maturity-model thinking into an original, evidence-weighted method.
Ready = corroborated evidence the capability is in place; Gap = corroborated evidence it is missing. Both count distinct scored KPIs, not repeats or one-off claims. A pillar with no Ready or Gap evidence is flagged thin.
Applied in this report: we run the framework on the Gen AI Leadership Forum as a worked example — its agenda and 175 datapoints are the evidence base. Each datapoint is read as Ready or Gap (which score), or as Vendor or Context (kept alongside as industry backdrop, not scored). The agenda is simply the example — the framework itself is agenda-independent.
How a single readiness score is built from the raw evidence — in the open. Every one of the 175 metrics is first sliced for the record, then read one by one; only the KPIs that measure the field’s own readiness feed the score, and each step is shown so nothing is taken on trust.
↓ Download this chart — Enterprise AI adoption & readiness funnel
The Forum's six agenda themes — Strategic Leadership, Hiring & Upskilling, Integrating AI into Core Systems, ROI-Driven AI, Gen AI vs Agentic, and Ethical AI & Governance — regroup into eight readiness pillars: the same 175 metrics, re-read as a maturity model.
Under each pillar's score: where the field is ahead, where it falls short, and the market momentum around it. This is a different lens from the headline “~20% AI-ready”: that figure counts how many organisations are ready; the maturity score weighs how strong the evidence of readiness is across all eight pillars — two views of the same gap, not competing numbers.
Pillar 1 of 8 · How ready is leadership for AI?
Strategy & Leadership
Related forum theme → Reshaping Strategic Leadership
Naturally light. This pillar draws less investment and market noise than others — strategic readiness is shaped more by how an organisation chooses to lead than by any tool or funding.
What this tells usThe signals here point to open challenges rather than finished wins — which makes strategic leadership the pillar with the most room to grow, and the clearest early-mover opportunity.
Pillar 2 of 8 · How ready are people & skills?
Talent & Culture
Related forum theme → Hiring & Upskilling
Heavy. Training vendors and L&D surveys generate a lot of noise here — expectation and marketing run ahead of the field's actual results.
What this tells usReal in-house capacity exists — but it’s concentrated in enterprise organisations (NBG, Cosmote), not the market — which is why it scores just 1.67. The opportunity is closing the engagement and talent-mix gap across the wider field.
Pillar 3 of 8 · Is the data & infrastructure ready?
Data & Infra Readiness
Related forum theme → Integrating AI into Core Systems
Enormous. This pillar is dominated by hardware and compute investment (GPUs, data-centres) — but that's the supply side building up, not organisations' own data readiness.
What this tells usThe field said little about its own data foundations — a quiet but important reminder that infrastructure readiness is the groundwork most organisations still need to lay.
Pillar 4 of 8 · Are AI use cases actually deployed?
Use Case Deployment
Related forum theme → Integrating AI into Core Systems
Moderate. Vendors showcase deployment playbooks and company scale — real momentum, but mostly supplier-led.
What this tells usThis is where AI is already delivering — the clearest proof that deployment works when organisations commit, and a template others can follow.
Pillar 5 of 8 · How widely is AI actually used?
Adoption & Usage Depth
Related forum theme → Reshaping Strategic Leadership · Gen AI vs Agentic AI
Light. Little money pools directly around adoption itself — it shows up as broad market activity rather than targeted investment.
What this tells usAdoption is real and accelerating globally; the opportunity is helping the local field catch up to where the market already is.
Pillar 6 of 8 · Is AI delivering business value?
Business Impact & ROI
Related forum theme → ROI-Driven AI
The loudest of all. This pillar is dominated by investment and funding figures — capital is pouring in (over $300B globally) well ahead of proven returns.
What this tells usInvestment is running well ahead of proven value — the opportunity is turning that capital into measurable results, which few have done yet.
Pillar 7 of 8 · Is AI governed safely & responsibly?
Governance, Risk & Trust
Related forum theme → Ethical AI & Governance
Heavy. A large regulatory and threat backdrop surrounds this pillar (GDPR, the EU AI Act, rising cyber-attacks) — pressure is high and working tools exist.
What this tells usThe foundations of good governance are appearing; the opportunity is scaling them to match the risks the field is already exposed to.
Pillar 8 of 8 · How fast is AI capability evolving?
Innovation Velocity
Related forum theme → Gen AI vs Agentic AI
Forward-looking. The activity here is projections and milestones (agent scaling, AI history) rather than money or current deployment.
What this tells usIt's early — this pillar is about trajectory more than today. The opportunity is being among the first to turn agent ambition into working reality.
Readiness averages 1.61 out of 5 across the eight pillars — an AI Delta of about 3.4 to a full score. The gap is uneven: Use Case Deployment is well ahead (4.17/5), while Strategy & Leadership, Data & Infra Readiness and Innovation Velocity sit at zero on the distinct scored KPIs the Forum reported. That spread is the readiness gap this report headlines — AI is already being deployed, but the strategy, data foundations and forward-capability behind it aren't yet backed by evidence of readiness.
The eight readiness pillars adapt established AI-maturity frameworks — Gartner's AI Maturity Model, Sema4.ai, Straive and McKinsey — organised into eight capabilities after the pattern of Nemko's AI-CMM. Two of them — Adoption & Usage Depth and Innovation Velocity — are PLH's own synthesis and the thinnest-sourced externally. Full per-pillar sourcing in Methodology → The maturity scores.
All 44 transformation KPI metrics from the Forum — the outcomes organisations are actually chasing with AI — mapped onto the eight readiness pillars. Position shows which pillar each metric flows into; colour shows what kind of evidence it is — the same four categories as the framework above.
Where AI transformation stands
Of the Forum’s 175 metrics, 44 describe transformation — the outcomes organisations are pursuing with AI. Mapped to the eight readiness pillars, they cluster tightly: two-thirds sit in Use Case Deployment and Business Impact & ROI, while adoption and innovation velocity register almost none.
That concentration is a natural starting order — enterprises put AI to work and prove its value first. The picture blends genuinely measured wins with the capabilities technology partners are bringing to market: in talent, governance and data the momentum is still largely supplier-led, and the strongest measured results so far come from a handful of organisations that committed early.
The takeaway: transformation today is real and encouraging, but concentrated — focused where it pays off fastest, and led by a few. The opportunity now is to broaden it from those early movers to the wider market, and into the areas — strategy, adoption, agentic velocity — that are still taking shape.
↓ Download this chart — Where enterprise AI transformation goes
The measured signal — Where AI is actually deployed, it works — Use Case Deployment scores 4.17 out of 5, the highest of any pillar. The problem is underneath: only about 1 in 5 organisations has the data, infrastructure and governance to scale it, and Data & Infra Readiness scores zero. Ambition isn’t the constraint — the foundation is.
The room’s read — The cost of NOT integrating is the backdrop that makes the case: offshore equipment failure $200K–500K/day, downtime $500K/hour, diagnostic lag 4–12 hours — and the only explicit rollout horizon anywhere: pilot 2–3mo, validation 2–4mo, fleet 6–12mo.
What it means for leaders — Fix the foundation before scaling — data readiness and governance, not more pilots. The deployment engine already delivers; what it stands on doesn’t yet.
The measured signal — Returns are real but small-scale: the one hard case is a €4.3K deployment that returned ~€100K (>2,000% ROI) — a single illustrative result, not enterprise-wide. Business Impact & ROI scores 2.50/5.
The room’s read — The money & attention around AI: aggregate investment $235B (2024) → $300–400B (2025), per-token inference cost roughly half last year’s (improving unit economics, not an overheating market), over a dense layer of VC fund mechanics — portfolio sizes, exit targets. Almost the entire theme is market backdrop.
What it means for leaders — Lower the internal threshold before approving pilots — ROI shows up at small scale first. Don’t wait for a large enterprise-wide business case.
The measured signal — Recovery is the gap: only 17% of organisations recover from a breach in <100 days, against a rising threat (+44% cyberattacks) — a widening exposure window. One governance practice is in place — a quarterly AI-Committee cadence — giving Governance, Risk & Trust its 2.50/5.
The room’s read — The backdrop: accuracy thresholds diverge — 95%+ for narrow healthcare AI vs ~80% in general software — and EU AI Act compliance is projected at ~€10B over 10 years, about half the EU’s separate ~€20B AI-infrastructure investment plan.
What it means for leaders — Treat governance as a continuous strategic function, not a quarterly checkpoint — the exposure window between breach and recovery is where risk compounds.
The measured signal — Volume is rising but quality lags: 63% of employees say training programmes need major improvement and 58% multitask during sessions (both Gaps). Europe’s talent structure is itself a gap — 3 lawyers per technical expert, the inverse of the US/China ratio. And where real in-house capacity exists, it’s concentrated in a couple of enterprise giants (NBG, Cosmote), not the market — so Talent & Culture scores just 1.67/5.
The room’s read — The backdrop: training completion is climbing — 41% (2023) → 50% (2025) of the global workforce — as L&D investment rises.
What it means for leaders — The talent-ratio imbalance is correctable through hiring policy directly — no culture-change programme required. Redesign upskilling delivery (format, relevance) before increasing budget or headcount.
The measured signal — Where AI is applied to internal work, gains sit in a steady band — +21% decision-making, +15% productivity, −40% downtime. But boards spend ~80% of their time on past-performance review, not strategy (a Gap), and Strategy & Leadership scores 0.00/5. What separates results is leadership commitment and programme design — not the technology.
The room’s read — (panel colour, not measured KPIs) The Forum’s Boardroom Panel put the split at roughly 5% active adopters, 5% who’ve written AI off, and ~90% still ‘preparing’ or ‘undecided’. The jaw-dropping numbers — like 50× more movie output — surface only as one-off stories.
What it means for leaders — The real gap is attention at the top: shift board time toward future-facing strategy and supply decisive sponsorship. Interest is abundant; commitment is scarce.
The measured signal — Real agent adoption is already here — a 40% agent-share of team composition in leading markets. The local target, at least one AI agent per development team by 2026, is a floor against that, not a stretch (a Gap). Innovation Velocity scores 0.00/5 on distinct scored KPIs; Adoption 2.00.
The room’s read — The headline speedups — 4,000×–40,000× — are narrow scientific simulations (wind-farm wake, physics models), compute-and-domain-specific. Treating them as a general productivity benchmark overstates what’s achievable in a typical process.
What it means for leaders — Engineering leadership here is ~one adoption cycle behind the leading markets — close it by embedding agents per team, and don’t benchmark against the 40,000× outliers.
The same 175 KPI metrics read through the lens of each room at the Forum.
The Forum's own Boardroom Panel put it starkly: ~5% of boards are active adopters, ~5% dismissive, and ~90% split between 'preparing' and 'undecided' — while boards spend ~80% of their time reviewing the past, not steering strategy. The constraint is decisive sponsorship, not interest or access.
The shift — approving AI as a line item → sponsoring the readiness agenda and sequencing pilot → scale
Where it goes — The opportunity cost of slow adoption is quantifiable, not speculative — the gap to the ~30% US comparator is the board's to close.
Readiness — data, infrastructure and governance — is the binding constraint: only 1 in 5 organisations are genuinely AI-ready. The cost of NOT integrating is the dataset's strongest ROI case: offshore downtime runs ~$500K/hour and diagnostic lag 4–12 hours.
The shift — running disconnected pilots → building the readiness stack and embedding agents per team
Where it goes — Engineering leadership here is roughly one adoption cycle behind the reference markets — closable, but only by fixing readiness first.
Where AI is deployed well, returns are large and fast — 40–95% gains recur across security, legal, industrial ops and training. But every case is a single, well-scoped deployment, not enterprise-wide transformation: strong at pilot level, unproven at full scale.
The shift — shipping one-off pilots → productionizing scoped wins into repeatable, measured ROI
Where it goes — Falling inference cost alongside rising investment signals improving economics — the window to scale scoped wins is opening, not closing.
Governance is one of the Forum’s biggest concerns — 17% of KPIs. It isn’t the largest theme (Integrating AI leads at 58), but it’s where the exposure window is widening fastest. The gap between threat growth (+44% cyberattacks) and recovery capability (only 17% recover in <100 days) quantifies a widening exposure window — yet governance is often framed as a quarterly checkpoint, not a continuous function.
The shift — periodic committee review → continuous governance aligned to EU AI Act tiering
Where it goes — As autonomy rises, trust and exposure management — not raw capability — become the market's dominant concern.
Training volume is rising faster than perceived quality: completion rose 41% (2023) → 50% (2025), yet 63% say programs need major improvement and 58% multitask during sessions. The structural signal is hiring composition: European startups run roughly the inverse of the US/China ratio of ~3 technical experts per lawyer. And where real in-house capacity exists, it’s concentrated in a few enterprise giants (NBG, Cosmote), not the market — which is why Talent readiness scores just 1.67/5.
The shift — generic AI-awareness training → rebalanced technical hiring and redesigned upskilling
Where it goes — The talent-ratio imbalance is correctable through hiring policy alone — no culture-change program required.
Capital is outpacing adoption — AI investment is growing ~10× faster than the economy ($235B → $300–400B) while Greek adoption sits at 9.5% against a ~30% US comparator. Falling per-token cost alongside rising aggregate investment signals improving unit economics, not an overheating market.
The shift — betting on model capability → backing readiness and technically-balanced teams
Where it goes — Improving unit economics plus an adoption gap = a market that is early, not late — for disciplined, readiness-focused capital.
Agentic AI — software that doesn’t just answer questions but actually does the work, taking action and finishing multi-step tasks on its own — is the frontier this series exists to track. It was also the least proven: only a handful of the 175 metrics touch on it, and just 7 are specifically about agents. Treat it as a sign of where things are heading, not where they are.
What’s real today — in the most advanced markets, AI agents already make up about 40% of a small software team, and a quarter of customer conversations are handled end-to-end by a bot. Locally, the ambition is set but not yet reached: one software company aims to put at least one AI agent per development team by 2026.
What’s still mostly promise — the tools claim big wins — up to 70% less manual security work, nearly half of routine analyst tasks automated — but these are supplier figures, not independently proven. And the boldest numbers — agents multiplying from a handful to hundreds per team within a year, software that rewrites its own code, billions flowing into AI-coding startups — are projections and demos, not evidence.
The honest read: the excitement is enormous, but the proof on the ground is thin — one 2026 target and a few supplier demos. That’s exactly why this is the frontier where organisations are furthest from ready, and where moving early matters most.
The verdict. The enterprise has started but isn't ready to scale: AI is live in most organisations, yet readiness — data, governance, talent — lags across the board. Scored across eight readiness pillars, the field averages just 1.61 out of 5 (an AI Delta of ~3.4 to target) — strong where AI is already deployed, near-zero on strategy, data foundations and forward capability. That gap, not access to the technology, is what the next five years turn on.
That gap breaks down into six conclusions — and each one is what a recommendation is built to answer.
Strategy & Leadership scores 0.00 out of 5 — the lowest pillar of all. Boards spend ~80% of their time reviewing the past, and roughly 90% are still ‘preparing’ or ‘undecided’. Interest is everywhere; decisive leadership is missing.
60–70% run AI live, but only 1 in 5 has the data and controls to scale it past a pilot; Data & Infra Readiness scores 0.00/5. The base can’t hold what’s being built on it.
Talent & Culture scores 1.67/5. Training volume is rising (41%→50%), yet 63% say programmes need major improvement and 58% multitask through them. Genuine in-house skill sits in a handful of giants, not the market.
Agentic AI is the least-measured area (Innovation Velocity 0.00/5, just 7 of 175 metrics) yet the fastest-moving. The 2026 target already trails leading markets, where agents make up ~40% of the team.
Governance scores 2.50/5 and runs on a quarterly cadence. Threats climb (+44% cyberattacks) while only 17% recover in under 100 days — the exposure window widens between checks.
Deployment is the strongest pillar (4.17/5) — but every win is a single, well-scoped pilot; large business cases stall, so nothing reaches enterprise scale.
Where each of those six is heading — external forecasts placed against the Forum’s trajectory. These are projections, not Forum measurements, and the Forum’s own agentic evidence is thin (just 7 of 175 metrics), so read them as direction, not proof.
Gartner expects 40%+ of agentic-AI projects to be cancelled by 2027, and MIT already finds 95% of GenAI pilots deliver no measurable return. Strategy and value, not technology, separate the survivors from the shakeout.
Gartner projects 33% of enterprise software will be agentic by 2028 (from <1% in 2024); McKinsey finds only about a third of firms have scaled AI, with data quality the top blocker. Only those that fix the base get there.
The WEF expects 39% of core skills to change by 2030; Korn Ferry projects 85M roles left unfilled for lack of skills. Upskilling is the gate to everything above.
Gartner projects agentic AI will autonomously resolve 80% of common customer-service issues (−30% cost) by 2029; McKinsey reports agent use is already the fastest-rising area of AI adoption. Mainstream — for those who build the bench now.
Gartner expects 40% of CIOs to demand ‘guardian agents’ by 2028 to autonomously oversee AI; IBM finds 63% of organisations still have no AI-governance policy. A quarterly check can’t cover that surface.
Gartner projects 15% of day-to-day work decisions made autonomously by 2028; MIT warns most pilots never leave the lab (only 5% reach real scale). Autonomy becomes the default — pilots that never scale get left behind.
This is where the findings turn into a to-do list. Six moves — each with who owns it, roughly when, one number to watch, and how to get there. Together they close the readiness gap the report measures.
The scarcest thing in the data isn’t technology — it’s leadership. Strategy & Leadership scores 0.00 out of 5, the lowest of every pillar: boards spend ~80% of their time reviewing the past, and roughly 90% are still ‘preparing’ or ‘undecided’ on AI. What’s missing isn’t interest — it’s leaders who can actually set and own an AI strategy. Build that at the top first; every other move here depends on it.
Most companies (60–70%) already use AI somewhere, but only 1 in 5 have the data and controls to run it properly. Adding more AI projects won’t help until that groundwork is fixed first.
AI readiness isn’t held back by the tools — it’s held back by people. Talent & Culture scores just 1.67 out of 5, and the gap is quality, not effort: training is rising (completion up from 41% in 2023 to 50% in 2025), yet 63% of employees say the programmes need major improvement and 58% multitask through them. Where genuine in-house AI capability exists, it sits inside a handful of enterprise giants, not the wider market. More of the same training won’t close this — role-specific upskilling that people actually apply will.
Agentic AI (AI that acts on its own, not just answers) is the least-proven area today but the fastest-growing. The 2026 goal — at least one AI agent per development team — is already behind leading markets, where agents make up ~40% of the team. Set up a dedicated skills-and-leadership programme now, before the gap widens.
Right now AI governance is a box ticked once a quarter, and boards spend most of their time looking backwards. Make it an ongoing job instead.
Where AI is actually switched on it works — the report’s strongest score (4.17/5) — but only as small, single-use pilots; big all-at-once business cases stall. So make small pilots easy to approve.
Do these six and the readiness score — a low 1.61 out of 5 today — starts to move: a fast start on weak foundations becomes something that lasts through 2026–2030. The gap is the plan.
