Skip to content
Based on Stanford AI Index, McKinsey, MIT, Gartner & Menlo Ventures (2025–26)

Everyone Adopted AI. Almost Nobody Is Banking the Value — Yet.

19 June 2026·11 min read

This is the first in a series of public research notes from the Fenrik Signals desk. Same method we run on a commissioned brief — pull the primary sources, separate the signal from the noise, end on what to actually do — pointed at the whole market instead of one company's.

The headline of 2026 is not that companies are adopting AI. They already have. The headline is the distance between adoption and return — and that distance is where the next two years of B2B revenue gets made.

TL;DR

1. Adoption is the solved problem

A year ago you could still win a meeting by explaining why a company should use AI. That door is closing. Stanford's AI Index 2025 put organizational AI use at 78% for 2024 — up from 55% the year before, a 23-point jump in twelve months. McKinsey's 2025 State of AI reads higher still: 88% using AI regularly, 79% using generative AI specifically.

The capital backs it up. Global corporate AI investment reached $252.3 billion in 2024, and Gartner expects worldwide AI spending to climb roughly 47% in 2026 to about $2.6 trillion. "Should we use AI" is no longer the question anyone is asking. So if your pitch still answers it, you're selling into a closed door.

2. The value gap is the whole opportunity

Here is the number that should reorganise how you sell.

Near-universal adoption, single-digit value capture. Companies bought the tools, rolled out the licenses, ran the pilots — and the money mostly hasn't shown up in the P&L. McKinsey's own framing for 2025 moved off the adoption headline and onto the scaling gap: only about a third of organizations say they've scaled AI across the business, and fewer than 10% are scaling AI agents in any function.

That gap is not a problem for your buyers to be embarrassed about. It is your addressable market. Every one of those companies has a budget line, an executive who championed the spend, and nothing yet to show a board. They don't need to be convinced to start. They need to be shown how to finish.

3. Why the pilots die

The most-cited number of the year explains the gap bluntly.

MIT's NANDA initiative — drawing on 150 executive interviews, 350 employee surveys, and 300 public deployments — found that 95% of generative-AI pilots produce no measurable P&L impact. Crucially, they trace the failure not to model quality but to what they call the learning gap: the inability to integrate AI into real workflows, structures, and incentives. The 5% that break through "pick one pain point, execute well, and partner smartly."

Gartner sees the same wall from the budget side, predicting that at least 30% of generative-AI projects would be abandoned after the proof of concept — citing poor data quality, weak risk controls, escalating cost, and unclear business value. None of those are model problems. All of them are integration and operating problems — which is to say, services problems.

The lesson for anyone selling AI work: the demo is not the deliverable. Production is. The firms winning right now aren't shipping cleverer models; they're doing the unglamorous integration the buyer's pilot skipped.

4. Where the budget is going in 2026

Spend is racing ahead of proof — and the shape of the spend tells you where to stand.

Menlo Ventures' 2025 State of Generative AI in the Enterprise — a survey of 495 US enterprise AI decision-makers — puts enterprise generative-AI spend at $37 billion in 2025, roughly triple the prior year and up from $1.7B in 2023. More than half went to applications. But the category everyone is talking about is barely funded: only about $750 million — roughly 10% of horizontal spend — went to standalone agentic platforms.

Meanwhile the forecasts are euphoric. Gartner expects spending on agentic AI in software to rise about 141% in 2026, overtaking chatbots and assistants by 2027 — while simultaneously warning that over 40% of agentic-AI projects will be cancelled by the end of 2027 for the same reasons pilots die today: cost, risk, and unclear value.

The read: agents are where attention and budget are heading, but the gap between "funded" and "working" is about to repeat one layer up the stack. The firm that can put an agent into production with guardrails and a measurable outcome — not a demo — gets to ride the wave without drowning in the 40% that get cancelled.

5. The market for closing the gap

This is a tailwind, not a niche. Independent forecasts put the AI-consulting services market at roughly $11 billion in 2025, growing toward $91 billion by 2035 — about a 26% CAGR (other houses model a similar curve from a slightly higher base). The demand exists precisely because the value gap exists: someone has to do the integration the software didn't.

The catch: the easy, top-of-funnel layer — awareness workshops, generic "AI strategy" decks, one-off pilots — is commoditising fast, and that's the layer the data shows failing. The defensible work is downstream: production, evaluation, workflow redesign, and measured outcomes.

6. What this means if you sell AI work

Pulling the five sources together, the desk's read for AI consultancies, training firms, and B2B operators:

  1. Stop selling adoption; start selling value capture. Your buyer already adopted. Lead with the EBIT gap — the ~6% capturing real value — and position as the bridge from "deployed" to "banked."
  2. Make "pilot-to-production" the offer, not a phase. The 95% failure rate is your wedge. A fixed-scope "production-readiness" engagement sells against a problem the buyer can already feel.
  3. Sell integration, because that's the actual blocker. MIT's learning gap and Gartner's abandonment data both point at workflow and operating-model work — not model selection. Price the unglamorous part; it's the part that pays.
  4. Be early on agents — but credible. With only ~$750M in real agentic spend and 40%+ of agent projects forecast to be cancelled, the opening is for whoever can ship one agent into production with guardrails and a number attached.
  5. Target the funded-but-stuck. The sharpest accounts are the 80–90% who adopted, have budget, and have nothing to show a board. They don't need a reason to start — they need a path to finish.

That last point is the whole business in a sentence, and it's exactly what this desk is built to find: not whether a market is moving, but which companies are stuck in the gap right now, and what would get them unstuck.

Sources

Figures are as reported by each source at time of writing (June 2026). Where reports phrase a statistic differently across summaries, we cite the conservative reading. This note is general market research, not advice on any specific company.

Share LinkedIn
#AI Adoption#Research#Commercial Strategy#Market Signals

Practical AI notes

One email when there’s something worth it — a new essay or field note, prompt templates you can paste in, a tool worth trying, or a build-your-own-agent walkthrough. No hype, no filler.