Accenture booked $2.2 billion in AI consulting last quarter. Here’s the part your engineering team could have handled for free.

Source: Nate’s Newsletter Date: 2026-03-24 URL: https://natesnewsletter.substack.com/p/youre-about-to-spend-millions-on

Summary

Accenture’s $2.2B AI consulting quarter is the headline, but Nate’s argument is about decomposition: of the five hard problems in agent deployment, four (context compression, codebase instrumentation, linting, multi-agent coordination) are within the reach of a competent engineering team using established software principles. Only the specification problem — translating business process into precise agent instructions — legitimately warrants domain expertise. The implication is that most of that $2.2B is fees for repackaged fundamentals, not novel expertise.

Implications

• The “specification problem as the only hard part” framing is a useful forcing function: if a consulting engagement isn’t centered on specification work, it’s probably replaceable with internal engineering time and open tooling.
• Feeds the enterprise AI adoption thread — the consulting revenue spike is a lagging indicator of early-adopter spend; as the specification problem gets better tooled (e.g., via structured prompt frameworks), the consulting moat compresses.
• Relevant for evaluating AI infrastructure vendors who position deployment complexity as a reason to buy managed services over building on raw APIs.