AI Strategy Consulting: What I Do and How It Works

Thomas Prommer Technology Executive & CTO Connect on LinkedIn

Published: May 28, 2026

Updated: May 31, 2026

Languages: Deutsch · Español · Français

Every week or two I get a version of the same call. A CEO or a board director at a growth-stage company has a problem they describe as "we need an AI strategy." When I ask what they have now, the picture is almost identical across companies. There is a steering committee that meets every other Thursday. There are three or four pilots running in different business units. There is a tier-one consulting deliverable on a shared drive that everyone references but nobody reads. The annual AI spend has crossed seven figures, and nobody can produce a clean slide that shows what came back for it.

What these companies do not need is more AI. They need someone whose job is saying yes and no to AI bets, with the authority to make the no stick. That work is what gets called AI strategy consulting, and the title hides more than it reveals. The practitioner version of the role and the analyst version share a vocabulary and almost nothing else.

I have spent twenty years running technology programs from inside Fortune 500 organizations, including the AI rollout at one of the largest sports apparel companies in the world. I have also done fractional CTO and CIO engagements at companies in the $50M to $500M range, which is where most of the AI strategy consulting market actually lives. This page is what I tell prospective clients before the first call, so the call can spend its time on their situation instead of on definitions.

What AI Strategy Consulting Actually Is

Strip away the deck templates and the workshop facilitation. The work comes down to four recurring decisions that someone has to own.

Which initiatives to fund, and which to kill

Most companies above $100M revenue have more AI ideas than they can afford to pursue seriously. The AI strategy consultant classifies every initiative as one of three things: a moonshot the company can afford to lose entirely, core operating leverage with a clear margin case, or table-stakes compliance work being done because not doing it would look strange. Each category is funded on different criteria and measured against a different bar. Initiatives that cannot earn their slot get cut. The cutting is the part that needs executive authority, which is why committees almost never do it.

Buy versus build, capability by capability

The fastest way to set money on fire in 2026 is to rebuild a capability that has a $99-per-seat SaaS equivalent already shipping. The slower, more expensive mistake is to buy a platform for something that should have been a competitive moat. The right answer is decided per capability, with a reasoning pattern the rest of the org can reuse the next time a similar question shows up. Harvard Business Review's work on managing generative AI risks argues the same point from the governance side: the decision is structural, not procurement-driven.

Org design and the actual talent mix

"We need to hire AI engineers" is almost always the wrong specification. The right one names the actual roles: applied ML engineers, MLOps infrastructure leads, prompt engineers, evaluation owners, vendor managers, AI product managers, and the decision-maker accountable per use case. The last one is the role most companies forget to staff. The AI strategy consultant maps which of those are in-house versus contract versus vendor-embedded, then writes the hiring plan that matches the portfolio.

Governance and the measurement cadence

Model cards, evaluation baselines written before the first model ships, kill criteria recorded somewhere the team cannot edit retroactively, unit economics per use case reported to the board on a fixed cadence. None of this is glamorous. None of it is where most of the AI spend goes. But without it, the funding decision in the first section runs on vibes instead of evidence, and the portfolio review turns into a political negotiation. NIST's AI Risk Management Framework is the closest thing to a public reference standard; most companies adopt a subset of it and customize the rest.

Practitioner Consultant vs. MBB / Big 4 Firm: The Real Choice

Most of the AI consulting market is dominated by the MBB strategy houses (McKinsey, BCG, Bain), the Big 4 (Deloitte, PwC, EY, KPMG), and Accenture. They are excellent at what they do. What they sell is team-shaped engagements: a partner who shows up for the kickoff and the readout, a manager who owns the project plan, two or three associates who do the research and build the deck. That model is appropriate when the company genuinely needs a parallel team to do discovery work in a domain the company has no internal capacity for.

The gap shows up at the decision layer. The partner who carries the relevant scar tissue is on the kickoff call and one of the readouts, and is otherwise running six other engagements. The day-to-day work, meaning the analysis and the deck-building, runs through associates and managers who are typically three to eight years out of school. They are smart and well-trained, and the framework they apply is reasonable. They are also not the person who has killed a $4M initiative because the unit economics did not survive the second 90-day review.

The practitioner consultant is the opposite shape. One senior person, sitting inside the executive team for one or two days a week, making decisions instead of writing about them. Less analytical breadth, far more decisional depth. The cost difference reflects the staffing difference: an MBB or Big 4 strategy engagement typically lands in the $400K–$1.2M range for 12 weeks of work; a practitioner-led equivalent runs roughly 25–35% of that for the same scope based on observed client engagements.

Neither model is universally correct. The right question is whether the company's bottleneck is at the analysis layer or the decision layer. For most mid-market companies past their first serious AI investment, the bottleneck is decisional.

Engagement Models I Offer

Three shapes, picked to match where the company actually is.

Advisor retainer, $5K to $15K per month

Quarterly deep reviews of the AI portfolio plus call-down access for the executive team in between. Appropriate when there is a competent internal AI leader who needs a sparring partner, when a board wants AI governance competence without adding a seat, or when a pre-Series-B company needs a sanity check on the next 18 months of decisions without a heavy engagement. Smallest commitment, lowest cost, and the engagement that produces the most value per dollar when the timing fits.

Fractional AI strategy executive, $15K to $40K per month

One to two days per week of senior executive time, sitting inside the company's executive forum, owning the portfolio decisions for 18 to 36 months. This is the engagement I run most often. It fits mid-market ($50M to $500M revenue) and well-funded growth-stage companies past product-market fit. The outcome is a portfolio with a single accountable owner, a measurement cadence the board trusts, and a permanent operating model handed off at the end. A fractional CAIO at this scale captures roughly 80% of a full-time Chief AI Officer's decision value at 10 to 20% of the fully-loaded cost.

Defined 90-day engagement, $75K to $250K

A scoped piece of work. Typically a portfolio audit, an org design, a vendor selection process, or a buy-versus-build decision on a specific capability. Useful when the company knows the question they need answered and wants a deliverable on a defined timeline rather than an ongoing relationship. The output is a written recommendation with the reasoning and the dissent surfaced, not a deck. Clients keep the document and use it for board reporting.

How to Evaluate an AI Consultant Before You Sign

Most of the bad outcomes I see start in the procurement process. The questions that filter well are specific.

Ask for a specific initiative they killed. Not a project that "wound down." A bet that was actively burning budget and was stopped on a specific date by a specific decision. If the answer is vague, they have not held the seat.

Ask to walk through a buy-versus-build decision they made in the last 18 months. The reasoning should include the cost model, the sensitivity analysis on the two or three numbers that drove it, and what would have flipped the decision. If the answer is framework-shaped instead of decision-shaped, they have written about the work without doing it.

Ask the smallest engagement they would take from you. Real practitioners say no to scopes that are too small or too large for the situation. Consulting firms with utilization targets cannot afford to.

Ask what would make them walk away. Anyone who cannot articulate the conditions under which they would resign from the engagement is going to oversell you.

Three Failure Modes I See Repeatedly

Hiring a technical AI leader for a strategy seat

Ex-research scientists, ex-ML infrastructure leads, and ex-AI startup founders are valuable people to have on the team. They are usually not the right shape for the strategy role. The job is capital allocation and org redesign; it shares a vocabulary with ML, but the work is different. A bias check: if the candidate's last two roles were "head of ML" or "distinguished AI engineer," they probably belong on the team, not in the strategy seat.

Splitting AI ownership across CTO, CIO, and CDO

"AI is everyone's job" produces committees, not outcomes. The hardest part of the role is killing initiatives that multiple stakeholders have reasons to keep alive. Without a single accountable owner, the kill decisions stall and pilots multiply. The visible symptom is a portfolio that has tripled in size over 18 months without a single graduation to production.

Buying a tier-one consulting deck when the bottleneck is decisional

A 60-slide deck with executive summaries and four-quadrant matrices is not an answer to the question of which three initiatives to fund this year. It is a description of the question, formatted for a steering committee. Companies that have already done the analysis usually need an operator, not another round of analysis.

Pricing Models for AI Consulting

The pricing conversation tells me more about a consultant than the pitch deck does. Four shapes cover almost every credible engagement I have seen, and each one solves for a different buyer problem.

Monthly retainer, fixed scope. The cleanest structure when the client wants predictable burn and a defined cadence. Advisor retainers run $5K to $15K per month for quarterly reviews plus call-down access. Fractional engagements run $15K to $40K per month for one to two days a week inside the company. The fixed monthly number forces both sides to define what good looks like before the first invoice, which is usually a healthy conversation.

Project fee, defined deliverable. A 90-day portfolio audit, a vendor selection, a buy-versus-build decision on a specific capability. Typical range $75K to $250K. The price is set by the scope of the question, not by hours. Project fees work when the company already knows the question. They fail when the company is paying for the consultant to figure out what to ask.

Day rate. Senior practitioner day rates sit between $2,500 and $7,500 in 2026, with the top of the range reserved for board-facing work and post-acquisition diligence. Day rates are honest pricing when the work is genuinely episodic. They become a procurement disaster when used to backfill a job that should have been a retainer. The signal that the structure is wrong: a day-rate engagement that has run for six straight months.

Contingent or equity-linked. Rare, and almost always a bad idea on AI strategy work. Equity in lieu of cash invites the consultant to optimize for the next funding round rather than the right portfolio decision. Success fees tied to specific cost-out targets occasionally make sense on procurement-heavy engagements, but the targets need to be set by someone other than the consultant. If a consultant offers to work for equity, ask why the cash conversation did not survive.

Two diagnostic questions before signing any pricing model. First, who pays for the time when the engagement scope changes (both directions). Second, what is the cancellation clause and the notice period. The first question separates real practitioners from consultants who treat scope changes as additional revenue. The second separates clients who are buying judgment from clients who are buying a contract.

Red Flags in AI Consulting Proposals

Five patterns recur in proposals from consultants who have not run the seat. The patterns are not subtle once you have seen them.

The discovery phase is more than a third of the engagement. A 12-week proposal where weeks one through five are discovery is a proposal where the consultant plans to learn the business on the client's dollar. Real practitioners come in with a hypothesis about where the money is being lost and validate or kill it inside the first two weeks. If discovery dominates the timeline, the consultant has not done this before.

Headcount is named before the work is named. A proposal that opens with a staffing pyramid (one partner, two managers, three associates, a researcher) is selling a team-shape, not a decision. The right shape of staffing falls out of the work. When the staffing precedes the work, the consultant is solving for utilization rather than for the client.

The deliverables are decks and workshops. A 60-slide final readout and three executive workshops is a description of activity, not of outcomes. The deliverable on a real engagement is a portfolio decision, a written recommendation with the dissent surfaced, or a hiring plan with named roles. If the proposal cannot articulate what changes in the client's operating model after the engagement, the engagement will not change it.

The case studies are anonymized to the point of meaninglessness. "A Fortune 500 financial services company" is not a case study, it is a paragraph. The version that signals real work names the problem, the decision, the dollar figure, and what would have happened without the intervention. The version that signals a re-skinned template lists industries and percentages without a concrete decision anywhere.

The proposal cannot say no. The strongest signal a consultant has held the seat is the question "what would make you walk away from this engagement." A consultant who cannot answer that question, or who answers it with a sentence about "alignment," will not say no to scope creep, will not say no to the client's pet project, and will not say no to the use case that should have been killed nine months ago. The capacity to say no is the product. A proposal that cannot demonstrate it is selling something else.

Where to Go Next

• What does an AI strategy consultant actually do? — the long-form answer to the category-defining question, with the engagement spectrum laid out end to end.
• AI consulting for small businesses — the five interventions that compound at SMB scale and the six that almost never do.
• Top AI consulting firms in 2026 — a practitioner's read on MBB, the Big 4, the boutiques, and the solo-operator tier.
• AI consultant salary — what the market actually pays at advisor, fractional, and full-time CAIO scale.
• AI consultant near me — the geographic angle on consulting: when it matters, when it doesn't, and what the LA market looks like.
• Enterprise AI strategy framework — the operating model behind the engagement: assessment, prioritization, platform.
• AI strategy executive: role and playbook — the organizational-design companion to this hub.
• AI leadership services — the direct path to scoping an engagement.
• Applied AI use cases — proof-of-work: real implementations across PR automation, agents, and training analytics.
• Technology executive: role, market, job — the executive-seat context for the consultant work on this page (working with a technology executive, or becoming one).
• Fractional CTO / CIO / Chief AI Officer — fractional AI consulting engagements when the work needs a seat, not a project.