Skip to content

The AI Adoption Gap

  • Title: The AI balloon is bursting and services must be ready to pick up the pieces
  • Authors: Saurabh Gupta and Phil Fersht (HFS Research principals; Fersht is founder and CEO)
  • Publication: Horses for Sources, the content arm of HFS Research, an enterprise-tech analyst and advisory firm
  • Date: June 25, 2026
  • Who it's really for: Enterprise leaders deciding where AI value comes from, and IT-services buyers weighing how to contract for it
  • Link: https://www.horsesforsources.com/wall-street-betting-on-an-ai-future-main-street-cannot-deliver_062526/

The one-line takeaway

Wall Street has written off the IT-services firms that enterprises actually need to make AI work. That gap is what the article is about.

The argument

The thesis. Wall Street and Main Street read AI in opposite ways. Investors have decided AI is the future and traditional services firms are the past. But the real state of enterprise adoption says those same firms are exactly who businesses need. The authors' framing: the market has mispriced what enterprises actually need.

The reasoning, step by step:

  1. Wall Street has rerated services as obsolete. The market prices AI-native firms at huge multiples and big services firms at a fraction. Accenture posted a solid quarter and the stock still fell hard. Good numbers, punished anyway, because the story turned against the category.

  2. Main Street's reality contradicts the story. On the ground, almost no one runs AI at scale. Most enterprises are still experimenting, most lack a coherent strategy, and most pilots never ship. The valuations assume an adoption that has not happened.

  3. The blocker is accumulated "enterprise debt." The real obstacle is not the models. It is a backlog of debt across technology, data, process, and talent. Messy systems and messy data are why pilots stall. AI cannot paper over that. Someone has to fix the foundations first.

  4. The firms Wall Street dismissed are the ones who can fix it. Resolving that debt takes deep knowledge of a client's systems, data, and processes. That is exactly what the big services firms already have. The market is writing off the capability that enterprises most need right now.

  5. But the old model has to change, toward "services-as-software." Billable hours compress. What replaces them: working down enterprise debt, and pricing on outcomes instead of labor. Firms also reach mid-market and emerging markets too costly to serve by hand, and expand past the CIO into the rest of the business.

The conclusion. Neither side wins alone. AI-native software cannot transform an enterprise by itself, and services firms cannot ignore AI. The two are converging. The authors' bottom line: the players who combine AI innovation with enterprise transformation capability are the ones who win the next decade. Wall Street's binary "AI versus services" bet misreads how that happens.

The evidence behind it

The market setup (public market facts the article reports, not original research):

  • Accenture's quarter: revenue up 6%, EPS up 9%, $3.6 billion in free cash flow, margins expanding. Yet the stock fell almost 18%, its worst single trading day on record. (Publicly checkable, but reported by HFS Research, 2026, which names no data provider.)
  • Leading IT services firms now trade at roughly 1.5x revenue, against an S&P 500 average of 3.7x. (Author's own analysis, HFS Research, 2026; inputs publicly checkable.)
  • AI-native firms command 20x to well over 100x revenue. Palantir sits around 60x, OpenAI around 35x, and newly public SpaceX close to 95x; the AI-native average price-to-sales is now above 45x. (Author's own analysis, HFS Research, 2026; inputs publicly checkable.)
  • Since their 2021 peak, the ten largest IT services firms have lost more than $600 billion in market value. (Author's own calculation, HFS Research, 2026.)

The adoption reality (the authors' own, HFS Research):

  • An estimated $18 trillion of accumulated technology, data, process, and talent debt across the Global 2000. (Author's own, HFS Research, 2026, from research across more than 2,000 Global 2000 leaders.)
  • Fewer than 1 in 10 enterprise AI pilots reach production. (Author's own, HFS Research, 2026; no third party credited.)
  • Only 13% of Global 2000 organizations have reached any meaningful level of AI maturity. The other 87% have not. (Author's own, HFS Research, 2026; the 87% is the arithmetic complement.)
  • 86% of enterprise leaders admit they lack a coherent AI strategy. (Author's own, HFS Research, 2026; no external source credited.)

The forecast (the authors' own, HFS Research):

  • Under the old model, combined IT services and software reaches roughly $5.5 trillion by 2035. Under services-as-software it compresses to around $4 trillion. (Author's own forecast, HFS Research, 2026.)
  • The enterprise prize from fixing the debt: about 8% faster revenue growth and a 16% cut in operating costs. (Author's own, HFS Research, 2026; no third party credited.)
  • A developed-market opportunity worth roughly $300 billion, plus an addressable market approaching $290 billion in emerging markets. (Author's own calculation, HFS Research, 2026.)
  • Services firms have historically operated across only 15% to 20% of enterprise spending, leaving the other 80% untouched. (Author's own, HFS Research, 2026.)

How much to trust it

Read this as an advisory firm arguing its own book. Carefully, not dismissively.

Who's talking, and what they sell. The authors are HFS Research principals. HFS sells research subscriptions and advisory services to the two groups the piece is about: enterprise buyers, and the services firms it defends. Sign-up and subscribe prompts bracket the article. No conflict-of-interest disclosure appears anywhere.

The thesis favors their business directly. The core claim is the most flattering possible story for HFS's own client base. It says Wall Street has undervalued services firms. And it says enterprises cannot deploy AI without the transformation expertise those firms provide. The piece frames the valuation gap as a market error to correct, which conveniently implies services should re-rate upward. That is a directional bet stated as analysis.

The big numbers are mostly their own. The load-bearing figures are HFS's, self-attributed and externally unverified: the $18T debt, the $4T and $5.5T forecasts, the maturity percentages stated with no methodology, and the 8% and 16% upside. The $18T is a constructed estimate of a composite, not an observed number.

What's actually independent. The public market facts are checkable: Accenture's 6% revenue, 9% EPS, $3.6B free cash flow, the roughly 18% drop, and the valuation multiples. But even these name no data provider, and the multiples trace to one author's own LinkedIn post. They are reusable only if you re-source them to filings and market data yourself.

Bottom line. Trust the direction of travel: debt is the bottleneck, adoption lags valuations, and integration still matters. Do not quote the HFS-original figures as if they were neutral measurement. Pressure-test or corroborate them before they land in your own deck.

So what — for you

  1. Make enterprise debt a board-level line item, not an IT problem. The blocker spans four buckets: technology, data, process, and talent. Audit which one is actually stalling your AI work. For most firms it is data and process, not the model. Budget for fixing it as a prerequisite to AI returns, not a side project.

  2. Stop celebrating pilots; demand outcomes on a clock. The authors' explicit bar is meaningful commercial impact within about 90 days, or you question continued funding. "We ran 40 proofs of concept" is a warning sign, not progress.

  3. Buy outcomes, not effort. Shift vendor contracts away from billable hours, seats, and token consumption, toward results tied to revenue, cost, or productivity. The authors claim fixing the debt unlocks roughly 8% faster revenue growth and a 16% cost cut. That is the kind of target to contract against.

  4. Pull AI above the CIO. The real spend and the real outcomes sit in core business functions. IT is only 15% to 20% of enterprise spend. Treat AI as a CEO and CFO operating-model question, not a procurement one.

  5. Don't over-rotate to AI-native point vendors. Vet a vendor on whether it can fix your foundations and ship to production, not on its AI story or its valuation halo.

The fine print

  • This is a vendor-position piece from HFS Research, with a thesis it sells ("services-as-software"). It carries no conflict-of-interest disclosure.
  • The big numbers are the authors' own research and forecasts, not independent third-party data. That includes the $18T debt, the 8% and 16% upside, the market-size forecasts, and the maturity percentages. Several carry no stated methodology.
  • The only figure attributed beyond HFS's own work is the AI-native valuation range, cited to a linked LinkedIn post by one of the authors. So it is still the same source.
  • The public market facts (Accenture's results, the valuation multiples, the $600B market-value loss) are checkable, but the article names no data provider. Re-source them yourself before relying on them.
  • Treat the directional argument as the durable takeaway. Treat the specific percentages as the authors' estimates to pressure-test, not facts to quote.

Sources

  • HFS Research (Saurabh Gupta and Phil Fersht) — The AI balloon is bursting and services must be ready to pick up the pieces, 2026. an estimated $18 trillion challenge; fewer than one in ten enterprise AI pilots make it into production; large IT-services firms trade at roughly 1.5x revenue while AI-native firms average above 45x. View source · verified 2026-06-30 · primary