The standard AI consulting engagement works like this. A company hires a strategy firm. The firm spends 3 to 6 months interviewing stakeholders, auditing processes, and benchmarking against competitors. They deliver a transformation roadmap. The engagement ends. Implementation is a separate contract, with a separate team, starting from scratch.
The result: McKinsey research found that 70% of large-scale transformation programs fail to meet their initial objectives. AI transformations are not exempt. The advice was sound. The execution never happened.
This model was built for a world where strategy and implementation required different skills, different timelines, and different organisations. That world is ending.
What AI has changed about the consulting timeline
Three years ago, building a custom AI system meant months of data preparation, model training, and infrastructure setup. It made sense to separate strategy from build. The build was expensive enough that you wanted a validated strategy before committing.
That is no longer true. Foundation models — GPT-4o, Claude Sonnet, Gemini 2.5 — have collapsed the time from 'we have a strategy' to 'we have a working system.' A competent team can build a proof of concept on a real business use case in 2 to 4 weeks. The PoC runs on your actual data, connects to your actual systems, and produces accuracy metrics you can measure.
When a PoC costs $10K to $25K and takes 4 weeks, the case for spending $200K and 6 months on strategy before building anything collapses. The build validates the strategy. The strategy without the build is hypothesis.
The new consulting model: advisory and build in one engagement
The model that is replacing traditional AI consulting has three phases, all in a single engagement.
Phase 1 is advisory. You get a build-vs-buy analysis, a technology recommendation, an ROI model, and a prioritised roadmap. This takes 1 to 2 weeks, not 6 months. The output is a specific, costed recommendation — not a general framework.
Phase 2 is the PoC. The same team that wrote the recommendation builds the working prototype. This is the critical structural difference. When the advisory team and the build team are the same people, the recommendation is constrained by what they know they can deliver. There is no knowledge transfer gap. There is no finger-pointing if the approach turns out to be wrong.
Phase 3 is the production build, if the PoC validates the approach. Fixed scope. Fixed price. You own the code.
Why traditional consulting firms struggle to make this shift
The advisory-then-build split was not accidental. It was a business model. Strategy partners billed at $400 to $800 per hour because they were scarce. Engineers billed at $200 to $350 because they were more available. Keeping them separate maximised revenue per engagement.
Large firms have internal incentives to maintain this separation. Changing it means partners sharing credit with engineers. It means fixed-price work instead of open-ended retainers. It means smaller but faster engagements that are harder to expand. These are not small changes. They threaten the firm's revenue model.
Specialist AI agencies do not have this conflict. Their incentive is a successful deployed system, because that is what generates referrals and follow-on work.
What this means for buyers: three things to demand from any AI consultant
If you are evaluating AI consulting firms, here is what the new model requires.
First: the advisory output should be a specific recommendation, not a framework. If the deliverable is a general 'AI maturity model' or a 'capability assessment,' you are getting strategy for its own sake. You want a costed recommendation on exactly which workflow to automate first, with which technology, and at what expected ROI.
Second: the same team should advise and build. If the firm hands off to a separate implementation team, the advisory team's accountability ends when they deliver the document. Insist on continuity.
Third: the exit condition for the engagement should be a deployed system with measurable results, not a slide deck or a roadmap document. If the consultant cannot tell you what will be live at the end of the engagement and what metric will show it is working, you are hiring a strategy firm — not an AI consulting partner.
Build-vs-buy and in-house-vs-agency: the two decisions that determine your AI path
The new AI consulting model requires answering two structural questions before any work starts. First: should you build a custom system or use an existing AI product? This is not always obvious — and the wrong answer costs 3x to 5x what the right answer would have. Our build-vs-buy framework maps this decision for the most common use cases.
Second: should you build an internal AI team or work with an agency? The honest answer depends on your headcount, your timeline, and how central AI is to your product strategy. We wrote a detailed breakdown of in-house vs agency that includes when each makes sense and what a hybrid looks like.
What the shift means for the consulting industry
The traditional AI consulting model will not disappear. There is still demand for strategy work at the board level, for M&A AI due diligence, and for regulatory compliance frameworks. Large firms will continue to win those mandates.
But the middle of the market — companies that want to automate a workflow, deploy a voice agent, build a RAG knowledge base, or integrate AI into their product — does not need a 6-month strategy project. It needs a 4-week PoC and a team that can take it to production.
That market is shifting to specialist firms that combine advisory with build. The strategy-only consulting model for AI will survive at the top. For everyone else, the new model has already won.
The Hestur model
Our AI consulting engagements start with a free 30-minute discovery call. We scope the use case, map the integrations, and give you a fixed-price recommendation by the end of the call. The PoC starts the following week. You have a working system — not a slide deck — within 4 weeks.