The AI Consultants Are Being Disrupted
Accenture just lost 20% in a day. Here's what that says about whether you still need a consultant
Last Thursday, Accenture's stock fell 20% in a single session. The company pointed to one cause: AI is disrupting consulting workflows. New bookings fell 2%. The firm that spent a decade selling AI transformation services to enterprises is now losing business because AI has started doing what its own teams used to do. This is the first public-market signal that the consulting procurement model is breaking.
In today's issue:
Main story: Accenture's bookings pause and what it signals about the consulting-or-build decision
Since Friday: Spotify discloses 73% of code PRs are AI-assisted, Cursor ships mobile agent supervision, California strikes a state-wide Claude deal, DeepSeek open-sources a software-only inference speedup, Andreessen joins the Defense Policy Board
Accenture's stock fell 20% last Thursday after the firm lowered its revenue forecast and pointed to one cause: AI is disrupting consulting workflows. New bookings fell 2%. The company that spent a decade selling AI transformation services to enterprises is now losing business because AI has started doing what Accenture's teams used to do.
This is not a technology story. It is a procurement story. Operators who are still planning to hire big consultants to lead their AI transformation should pay attention to what just changed.
The standard read is that Accenture is being disrupted by the very tools it sold. The irony gets a lot of coverage. What gets less coverage is the specific mechanism: clients are pausing new bookings, not canceling existing projects. That pause is meaningful. Companies are not abandoning AI consulting. They are waiting to figure out whether they need the consulting at all, or whether AI will just do the knowledge work the consultants were going to do.
That pause is the turn. The decision of whether to hire a big consulting firm has gotten genuinely harder, and clients are starting to feel it in a way that shows up in bookings.
The model behind Accenture's business is that large enterprises do not have the internal expertise to design AI strategy, so they buy it externally. Accenture, McKinsey, and Deloitte sell senior expertise at $400 to $600 an hour and then execute with junior staff at $100 an hour. The margin lives in the gap. Clients accept this because the alternative (building internal expertise) feels slower and riskier.
What is changing now is that the calculation for building internal expertise has shifted. AI can compress months of research and strategy design into days. A COO who would have hired a $500K consulting engagement to produce an AI strategy now has access to tools that can draft a credible version of that strategy in an afternoon. The tools are not perfect. But they are good enough to raise the question of whether the $500K was ever buying expertise or just buying time and confidence.
The Accenture drop is the first public-market signal that large enterprise clients are asking that question out loud.
The California deal announced the same week is worth reading alongside the Accenture news. California signed a first-of-its-kind agreement giving every state agency, city, and county access to Claude at a 50% discount. Every government employee in the largest US state now has an AI productivity tool available. California is not a cutting-edge technology buyer. If they have moved, the procurement curve has moved.
Anthropic's positioning here is deliberate. They are establishing Claude as the government standard through aggressive pricing, the same logic a platform uses to build distribution before competition arrives. The race to become the default AI layer at large institutions is happening right now, and it is being won through pricing and procurement relationships, not benchmarks.
The Anthropic-Amazon renegotiation reported by The Information this week adds another dimension. Anthropic has shifted its Amazon partnership from compute-hour pricing to per-token pricing. As Claude embeds deeper into Amazon's shopping, coding, and workplace tools, the model provider gained enough leverage to demand better terms. Amazon now pays more as usage scales.
When the original deal was struck, Anthropic needed Amazon. Now the dependency has partially reversed. Model providers who get embedded in production workflows will renegotiate from a position of strength. Cloud providers who assumed they controlled AI distribution because they controlled compute are discovering that distribution sits at the model layer, not the infrastructure layer. For operators who have built on a single cloud's AI offerings, the pricing and terms of that relationship are not fixed.
TEAM READINESS AUDIT
Get your team’s AI readiness score. Leave with an AI readiness brief.
A 15-minute assessment that turns your answers into a real decision artifact: your readiness level, your six-axis shape, where you're strongest, where you're constrained, and what not to build yet.
Counterargument
Accenture's 20% drop is being read too broadly. The firm has unique exposure: its AI revenue mix is higher than most peer consultancies, and some of the bookings pause reflects clients waiting for new Accenture AI offerings rather than departing entirely. McKinsey and BCG have not reported similar pressure, and there is a version of this where Accenture's pain reflects a transition, not a terminal decline.
My honest read: the transition version might be right for the large firms with deep institutional relationships, but it is less right for the $500K-to-$2M mid-market consulting engagements that most companies in your revenue band actually buy. Those engagements are sized for exactly the kind of work AI can now plausibly do, without the C-suite access and political relationships that make the $5M McKinsey contracts resilient to this particular pressure.
What this means for you
If you have been holding off on building internal AI expertise because you planned to hire someone to figure it out, this is the week to revisit that assumption. Not because consulting is dead, but because the gap between what you can build internally and what you used to need to buy is closing faster than most companies have updated their plans for.
A few things worth sitting with:
The next AI consulting pitch you receive, ask what portion of the deliverable could be produced by the same tools your team already has access to. Not as a gotcha, but as a genuine calibration. The answer will tell you whether you are buying expertise or buying comfort.
If your organization is on AWS or Azure, your pricing relationship with the AI models embedded in those platforms may not be as stable as you assumed. The Anthropic-Amazon renegotiation is not an anomaly. It is a preview.
Watch California's implementation closely over the next 90 days. Not because state government is a model for your org, but because how 400,000 government employees actually use a productivity AI tool (what they use it for, where it fails, what governance questions surface) will produce more real-world data about mid-market AI deployment than any whitepaper Accenture publishes this year.
From the field
I have been in a lot of conversations this month with operators who are going through their first serious AI procurement decision, not piloting anymore but actually buying. And the question I keep hearing is some version of: "Do we hire someone to help us figure this out, or do we figure it out ourselves?"
Two years ago, most of them hired someone because the internal capability was not there. Now I am seeing something different: the teams that have been running small pilots have built more genuine understanding than they realize. They know what works and what does not in their specific environment. What they lack is not expertise. It is confidence that what they know is enough to act on.
The Accenture number is partly telling that story. The clients who paused new bookings did not stop caring about AI. They stopped feeling like they needed someone else to think about it for them. That shift, when it happens inside an organization, is the actual inflection point.
SPONSORED BY CLUTCH
Hire secure AI teammates that work 24/7.
Hire pre-built AI teammates. Give your engineers and operators a platform to ship their own AI apps. Stop losing sleep about what is running where.
Clutch is the platform behind both: pre-built agents for the workflows your ops team should automate first, plus the integration plane your team's vibe-coded apps and Claude Code projects plug into. One platform. Real production. Visible and safe by default.
Built for ops, engineering, and security teams that are tired of the shadow-AI surface area inside their own company.
SINCE FRIDAY
Spotify's VP of Engineering disclosed that 73% of the company's code PRs are now AI-assisted, with 4,500 production deploys happening daily. That is a real engineering org at scale, not a startup. If you are benchmarking your own team's AI adoption, Spotify is now the reference point.
Cursor shipped an iOS app that lets developers launch and monitor coding agents from their phones. The obvious read is a developer productivity tool. The less obvious read is that agent supervision is becoming a mobile-first workflow, which has implications for how you think about oversight and approval processes.
California signed a deal giving every state agency access to Claude at a 50% discount, making it the first AI productivity tool deployed across an entire state government. The procurement curve has moved well past early adopters.
DeepSeek open-sourced DSpark, a software-only inference optimization that delivers 85% faster performance on existing hardware. The implication for hardware export controls as a strategy for containing Chinese AI progress is straightforward: software cannot be embargoed the same way chips can.
Marc Andreessen was appointed to the Defense Policy Board, officially putting a16z in an advisory role on Pentagon AI and national security decisions. Commercial AI in defense applications is about to move faster.
P.S. If you are in the middle of an AI procurement decision right now (whether to hire externally or build internally), reply to this email and tell me where you are stuck. I read every reply, and this is exactly the kind of question we work through at Seeko.
Haroon