Generative AI Strategy Consulting

Generative AI strategy consulting is advisory work that decides where generative capability changes a company’s economics, in what order, and on what evidence. The output is a small number of funded moves tied to demand the business intends to own, an operating model that names who decides what, a position on model choice, and a line from each investment to a number the board already tracks. Vivaldi has built demand-side prioritization since 1999. We advise on generative AI strategy, and we are explicit about the limit: we do not build, train or run models.

• Since 1999demand, commercial and customer strategy
• 200+published case studies
• 10 officeson four continents
• No vendor tiesour model advice carries no commercial stake
• The Interaction Fieldwritten by our founder

When companies bring us in

The call usually comes after the pilots worked. The technology is not the problem, and that is exactly what makes the next budget hard to defend.

• A funded program of twenty or more initiatives with no defensible reason for the order they sit in
• A board that has asked what the company’s generative AI position is, and an honest answer that is a list of tools
• Pilots that succeeded on their own terms and never graduated, because no operating budget was willing to adopt them
• A CFO asking why year two should cost more than year one, answered so far in adoption rates and hours saved
• An architecture quietly standardized on one model provider, with no view on what happens when the frontier moves

Why the use-case inventory does not work

The reasonable version of the standard approach is an inventory. Gather the ideas from across the business, score each for feasibility and value, fund the top of the list. It is orderly, it involves the right people, and it produces a portfolio nobody can object to.

It is also supply-led. It starts from what the technology can do and goes looking for somewhere to apply it, which ranks opportunities by enthusiasm and technical tractability rather than by economic consequence. The program that results is busy, individually defensible and disconnected from the numbers the board watches. Most executives never see the results the business case promised, and the cause is almost never the model.

Three failures recur underneath it. Value is counted in activity, because token volumes and time saved are measurable and economic profit is not measured by anyone in the room. Governance is written as a blanket approval rule rather than an escalation threshold, so every decision queues behind the few that are genuinely consequential. And where the savings are real, they are competitively symmetric: the same capability is on sale to your competitors at the same price, so efficiency gained this way is handed to customers within a cycle or two. The question that survives a budget review is not what generative AI can do for you, it is who ends up keeping the value generative AI creates.

What our generative AI strategy consulting services include

1. A demand-side map of where generative AI changes the economics

We start from where future demand will pool and where you have a right to win, then ask which decisions and workflows stand between you and it, and which of those generative capability actually changes. That ordering is the whole argument of the page. A map built this way rules initiatives out, which is the part an inventory never does.

2. A prioritized shortlist, with the rejected use cases and the reasons

A short list of moves, costed, sequenced, with the reasoning written down and the rejected candidates listed beside it. The rejections are the useful artifact: they are what stops the same idea being refunded in the next planning round under a new name.

3. An operating model that lets work graduate

Pilot purgatory is rarely a technical state. It is an ownership gap. We name who is accountable for the outcome once a prototype works, which budget carries it, and how funding moves from experiment to production without a fresh business case each time. Governance is designed to sit inside the workflow with an explicit escalation threshold, set on materiality, reversibility, precedent, exposure and the system’s own confidence, rather than beside the workflow as a blanket approval that slows everything to protect against a handful of edge cases.

4. A position on model choice and portability

Model selection is a strategic decision, not a procurement one. Capability, cost and terms move quarterly, and an architecture that assumes one provider is a bet that one company stays ahead on every dimension indefinitely. Nobody has held that for more than a few quarters. We specify models per workload with the choice designed to be reversible, knowledge and data held in your architecture rather than in a provider’s format, evaluation run on your own tasks and edge cases rather than public benchmarks, and a switching path that can actually be exercised. We hold no partnership with any model provider, so the recommendation carries no commercial interest either way.

5. Measurement in the units the board already uses

Every proposed investment is traced through to the spread between return on invested capital and the cost of capital: which demand space it serves, what right to win it rests on, then its effect on price, volume or capital intensity, then economic profit. If an initiative cannot be placed on that line, we say so before you fund it. Adoption rates, token volumes and hours saved are activity measures, and they are the reason so many AI programs cannot survive a budget review.

How a generative AI strategy engagement works

1. Frame. Interviews with the people who own the affected decisions, an inventory of what is already funded and what it was promised to deliver, and a read of where demand is moving in your category.
2. Prioritize. Each candidate tested against the demand space, the right to win and the route to economic profit. The output is a sequenced and costed shortlist plus a written list of what we would not fund and why.
3. Design the operating model. Ownership, funding route from pilot to production, the escalation threshold, model placement criteria and an evaluation approach built on your own tasks.
4. Instrument. A measurement line from each funded move to the board’s existing numbers, and the review cadence that decides what gets refunded, stopped or handed to an operating budget.

Why companies choose Vivaldi for generative AI strategy

Three reasons, and they are all structural rather than claims about effort.

We are a strategy firm, so the question we are equipped to answer is which moves are worth making, not how to ship them. We have no partnership with a model provider or a platform, so there is no version of our advice that is also a sales motion. And the prioritization method is not new to us: deciding what a company has permission to win and funding the few moves that follow is the work we have done since 1999, across 200+ published case studies. Generative AI changes the inputs to that question. It does not change the question.

How to evaluate a generative AI strategy consultant

Worth asking of every firm on your shortlist, including us.

• Who are you partnered with? A firm with a commercial relationship to a model provider or a platform is not a neutral advisor on which one you should use. That is structure, not dishonesty. Ask, and weigh the answer.
• Describe a use case you advised a client against. A firm that has never talked a client out of an AI investment has either been very lucky or has not been advising.
• How does this reach economic profit? Ask them to draw the line from a proposed initiative to the spread between return on capital and cost of capital. If the answer arrives in efficiency language only, you are buying cost reduction your competitors can buy too.
• What happens when the model changes? Ask what breaks on the next release and how you would find out. No evaluation and no regression path means something you cannot maintain.
• Who owns this in twelve months? Name the person inside your organization. If the honest answer is nobody, the engagement is selling delivery rather than a capability.

Strategy consultancy, AI implementation partner or model vendor?

Where our work stops and build work starts

We are a strategy firm. We do not build, train or fine-tune models, we do not write retrieval systems or production prompts, and we do not run deployments. None of our 200+ published case studies is a generative AI build, and we are not going to present adjacent work as if it were. What we bring is the decision in front of the build: which moves are worth making, in what order, who owns them, how model choice is kept reversible, and how each one reaches a number the board already watches. Implementation partners and model vendors do the rest, and our recommendation on which of them you use carries no commercial stake.

Related work

• AI business model innovation strategy if the question is which assumptions in your business model were built around costs that AI is now collapsing.
• Machine learning, predictive analytics and AI strategy where the value is in prediction, optimization or classification rather than generation.
• Strategic AI transformation if the program is enterprise-wide and the constraint is the transformation rather than the generative AI question itself.
• Demand landscape mapping if the demand architecture this prioritization depends on does not exist yet. Map that first.
• AI innovation capability consulting if the gap is what your teams can do rather than which moves to fund.

This work sits inside our Competitive Strategy group, part of Business Strategy and Transformation at Vivaldi, the strategy firm behind the Demand Landscape and the Interaction Field.

Generative AI strategy consulting in the US and worldwide

Our AI strategy work runs out of ten offices on four continents, with the US team in New York. Prioritization does not transfer cleanly between markets: the same capability lands on a different competitive structure, a different regulatory posture on automated decisions, and a different set of alternatives available to the customer. We scope by market and reuse the method rather than the shortlist.