AI-Geschäftsmodell-Innovationsstrategie

AI business model innovation is the work of finding which assumptions in your current business model were priced around costs that AI is now collapsing, and deciding what becomes viable when those costs fall. It is not a use case list and it is not a technology selection. It asks a narrower question than business model work in general: what can this company now be paid for, by whom, and on what unit, that it could not credibly sell eighteen months ago. Vivaldi has worked on growth and business model questions since 1999.

• Since 1999growth strategy and business model work
• Über 200veröffentlichte Fallstudien
• 10 Niederlassungenauf vier Kontinenten
• 5xWirtschaftsWoche: Das Beste aus den Bereichen Beratung, Strategie und Innovation

Wenn Unternehmen uns beauftragen

• AI pilots are running in six places, each one is defensible, and nobody can explain what any of them changes about how the company earns money.
• A cost the pricing model was built around has moved, usually the cost of producing something bespoke at volume, and the current price list no longer reflects what the work costs to deliver.
• A competitor or a new entrant has started selling an outcome where the category used to sell an input, and the sales team is being asked to defend a unit of value the customer has stopped caring about.
• The board has approved AI spend and asked what the return is, and the honest answer so far is a set of efficiency numbers that the market will read as harvesting rather than as growth.
• Two or three AI-enabled model options are on the table, usually a subscription, an outcome-based fee and a data or platform play, and nobody can say which one customers would let this company run.

Why the AI use case inventory does not produce a new business model

The standard sequence is familiar. Workshops across functions, a long list of candidate use cases, a scoring exercise on feasibility and impact, a heat map, a prioritized roadmap. It is a reasonable way to allocate an automation budget and it almost never changes a business model, because every input came from the business as it already runs. Scoring ideas generated inside the current structure returns a faster version of that structure.

There is a second problem, and it is financial rather than strategic. Most AI work on that list is supply side: fewer hours per output, fewer people per process, lower cost to serve. That work is often correct and worth funding, and all of it signals extraction, because every part of it is about taking more out of demand that already exists. A demand side change signals something different, that the earnings stream extends rather than tapers. Same spend, different story, different multiple.

The order that changes the answer runs the other way. Start from what customers want and cannot get, identify which of those things were previously uneconomic to deliver, and check whether the cost that made them uneconomic is one AI has actually moved. Then design how you get paid for it. Most AI capability is available to your competitors on the same terms and at roughly the same price, so capability on its own is not an advantage. What is defensible is the model built on top of it: a pricing unit, a participation structure or a data position that compounds as more customers use it.

So the test we apply at the end is blunt. If the company is not able to sell something it could not sell before, or charge on a basis it could not charge on before, the work was an efficiency program with a new name, and it should be taken to the board as one.

What our AI business model innovation services include

Clients arrive asking for different things: an AI strategy, an AI business case, a new revenue stream, a view on what AI does to their category. The work below is the same in each case, and it is sequenced so the expensive parts only get funded once the cheap parts have survived.

1. Model assumption audit: which costs did AI actually move?

We take the current model apart into the assumptions that hold its economics together: what makes a unit expensive to produce, what makes a customer expensive to serve, what makes a segment too small to address, what forces a human into the loop. Then we mark which of those have genuinely changed and by how much. Several will not have changed at all, and saying so early is the cheapest part of the engagement. This is also where most AI business cases quietly fail, because the cost they assume will collapse is not the cost that was binding.

2. Demand that was previously uneconomic to serve

The interesting territory is the demand your company already understood and declined to serve, because the economics did not work: segments too small for a bespoke offer, questions too expensive to answer individually, a service too labor intensive to price. We research that demand directly with customers and non-customers rather than inferring it from the use case list, because an inventory of what the company could automate says nothing about what anyone wants to buy.

3. Value capture redesign: the unit, not the price

Two questions, usually answered as one and usually answered wrong. What value does the company create under the new conditions, and separately, how is it paid for that value. When the cost of producing an output falls toward zero, pricing per output stops making sense, and the decision is which new unit is defensible: a seat, an outcome, a resolved case, a share of the value created, a subscription to a standing capability. Most AI business cases change what the company does and leave the invoice exactly as it was.

4. What compounds, and what any competitor can buy

Models that hold their advantage create value between participants rather than deliver it to them. We separate the parts of the proposed model that get stronger as more customers, partners and complementors use it, usually a proprietary data position or a network of participants, from the parts that are a purchased capability any competitor can acquire on the same terms. If the hundredth customer makes the offer no more valuable than the tenth, you have a new product rather than a new model.

5. Permission testing with customers

Customers decide what a company is allowed to become, and AI narrows that permission in places rather than widening it. A model your buyers do not believe you have the right to run fails whatever the economics say, and the question is sharper here than elsewhere: whether this company is trusted to make a given decision by machine, what it is allowed to do with the data the model depends on, and what has to stay visibly human. We test that with the customers the new model would depend on, before the capability build is funded.

6. The economic profit case and the transition

Every candidate model is run through the same arithmetic: what it does to price, to volume, and to the capital it consumes, and therefore whether it earns a return above the cost of that capital. A model that moves none of the three has not changed the business. We then design the sequence, since a new model runs alongside the old one for longer than anyone plans for: what gets piloted, what is ring fenced from the existing P&L, which metrics the new model may be judged by while it is small, and who owns the decision to scale or stop.

How an AI business model innovation engagement works

1. Frame. Leadership and CFO interviews, a read of where economic profit is earned today, and the current model written down as a set of explicit cost and constraint assumptions. The output is a short list of assumptions worth testing against what AI has actually changed.
2. Design. Two or three model options built out properly: who pays, for what unit, through which participants, what data position sits underneath, and what capital the model consumes. Each option carries its own economics rather than a shared one.
3. Test. Permission tested with customers, economics stress tested with finance, and the technical assumptions checked against what is reliably deliverable today rather than what is demonstrable once. Options die here, which is the point.
4. Commit. One model, a transition sequence, the metrics it will be judged by while it is small, and the governance that decides when it scales or stops. We hand over a decision with its exclusions written down.

Why companies choose Vivaldi for AI business model innovation

We treat this as a capital allocation question rather than a technology question, which means the work is built to survive a CFO reading it. We do not build or resell AI systems and we are not partnered with any model or platform provider, so the recommendation to do less carries no commercial cost for us. In practice that recommendation is common: for many companies the honest answer is that AI changes the cost line and not the model, and the valuable output is knowing which half you are in before the capability budget is committed.

The brand question and the model question are handled by the same team, because what customers will let a company become is a hard constraint on what the model can be, not a communications problem to solve afterward. That link is also where the financial argument comes from. Our 2026 study of twenty-four years of stock market data, Unrealized Value, found that companies with strong brands beat the S&P 500 by 3.7 percentage points a year between 2001 and 2024, turning a dollar invested into $15.85 against $7.03 in the index. The market pays a low multiple for a harvesting story and a high one for a growth story, and how a company chooses to earn from AI is one of the clearest places that difference now gets decided.

Vivaldi has been doing business model work since 1999, with 10 offices on four continents and more than 200 published case studies, and the firm has been recognized five times in the WirtschaftsWoche Best of Consulting awards for strategy and innovation.

One thing this page does not do is claim published AI business model case studies. No case study in our library is tagged to this service, and the work that is published under neighboring services was done before generative models were a factor. Where you want to see proof, ask us in the room for the relevant engagement rather than reading a borrowed case here.

How to evaluate an AI business model innovation consultant

Das sollten Sie bei jedem Unternehmen auf Ihrer Auswahlliste erfragen – auch bei uns.

• Ask which cost they think AI has moved in your business, before they propose anything. A firm that cannot name the binding constraint in your current economics is going to sell you a roadmap built from somebody else’s.
• Ask what changes on the invoice. If the proposal changes what the company does and leaves the pricing unit untouched, it is an efficiency program. That may be the right answer, and it should be called by its name.
• Ask what competitors can buy on the same terms. Most AI capability is purchasable. If nothing in the proposed model compounds, the advantage lasts as long as it takes a competitor to sign the same contract.
• Ask how they test customer permission. Automated decisions and data use narrow what a brand is allowed to do. A firm with no method for testing that is designing in a vacuum.
• Ask who they are partnered with. A firm with a reseller or implementation relationship behind the recommendation is giving you a procurement answer, and the proposal will rarely be to build less.

AI business model innovation consultancy vs AI implementation partner vs large management consultancy?

Verwandte Arbeiten

This page answers a narrow question: what AI changes about how your company creates and captures value. Several neighboring Vivaldi services answer adjacent questions, and two of them are the right destination often enough that the boundaries are worth stating plainly.

• Business Model Innovation Consulting is the canonical page for business model work and the right starting point if AI is not actually the reason the model is in question. Start there, not here, when the trigger is stalled growth, a category reorganizing or a pricing logic that stopped working for reasons that predate AI.
• Generative AI Strategy decides which generative moves to fund and in what order, including applications built on large language models. That is a different question from what AI does to the model you earn from, and it is answered there rather than here.
• New Business Model Design is the build rather than the diagnosis, with a platform and ecosystem emphasis, once it is settled that the model has to change.
• Business Experimentation tests a single assumption cheaply before capital is committed to it, which is a method used inside this decision rather than the decision itself.
• Machine Learning, Predictive Analytics and AI Strategy is where the question is what a prediction is worth inside a recurring commercial decision.
• AI Agents and Customer Personalization Strategy covers what changes when systems act rather than recommend.
• Strategic AI Transformation is the enterprise-wide program view, across functions rather than across the model.
• AI Innovation Capability Consulting builds the internal capability to keep doing this work without us.

All of it sits inside our AI Strategy and New Tech practice at Vivaldi, the strategy firm behind the Demand Landscape and the Interaction Field.

AI business model innovation in the US and worldwide

We run this work from 10 offices on four continents, with US, European and Asian teams on the same engagement where the model crosses markets, which it usually does. The economics travel further than the permission does. What a customer will accept being decided by a machine, what a company may do with the data a model depends on, and what has to stay visibly human differ enough between the US, Europe and Asia that a model validated in one market can fail in another for reasons that have nothing to do with the idea or the technology. European data and AI rules also bind the model design itself rather than only its implementation, so where a model is intended to travel we test permission and economics in each major market before the transition plan is written.