Your pilot portfolio is growing and nothing is in production

• The symptom you can see without a diagnostic
• The usual explanation, and why it does not hold
• What actually failed: predictability at the handover
• How to check this in your own organisation
• The structural response
• Conclusion
• References

Validated pilots stall at the moment of handover, and the gap sits in who owns the work once the pilot is over.

The symptom you can see without a diagnostic

The slide comes up in the quarterly review. Active pilots: nineteen, up from fourteen. Someone asks how many are running in production. The answer is one, from two years ago, and it went through because a business unit director happened to want it badly enough to pay for it out of her own line. Nobody in the room disputes the numbers. Nobody has an explanation either, so the conversation moves on to the next slide.

Two records make the pattern visible, and neither of them needs a consultant to read. The first is the count of pilots started, which almost always rises year on year. The second is the list of solutions that are in operational use with a budget line, an owner, and a support arrangement, which in most organisations is short and grows slowly. When the first record moves and the second does not, the organisation is producing validated pilots as an output in itself.

Across the interviews for our second study this came up as a recognised condition with a shared vocabulary. Corporate innovation leaders described pilots that met their own success criteria and then stopped.

“Even successful pilots don’t automatically have a place in the organization.”
Strategy Director · Telecommunications · Denmark

The phrasing is worth reading closely. The problem is not the pilot’s result. It is the absence of a place for the result to go, and the absence of anyone whose job it is to make one.

The usual explanation, and why it does not hold

The standard account of a growing pilot portfolio with no production output blames the entry stage. On this reading, the organisation picked the wrong pilots, the technologies were immature, and the remedy is a tighter funnel: better screening, a scoring model, fewer and more serious experiments. It is a comfortable explanation because it puts the fix in a place where the innovation function already has authority.

The data does not support it. In our interviews the solutions that stalled were, in the main, ones that had already passed their pilot criteria. They were evaluated as relevant, they worked in the test environment, and they stopped anyway. A screening model applied at the entry stage cannot correct a failure that occurs after validation, because by then the screening has already done its job.

What changes after the pilot is the evaluation logic. Innovation and transformation teams read a successful pilot as evidence for continuation. Operational units read the same pilot through continuity, process fit, resource allocation and long-term maintainability, and they apply that reading for the first time at the point where the pilot is already over.

“What works in a pilot does not always fit into existing processes.”
Operations Director · Logistics · Sweden

A second shift happens at the same moment. Procurement, legal, intellectual property, cyber and data-security review become structural gates between pilot validation and operational adoption. For the external partner this is a change of subject: the task moves from proving that the solution is relevant to proving that it is reliable, supportable and compliant over years. Neither side is usually told that the criteria have changed, so both continue to argue the case they prepared for the previous stage.

External evidence points the same way and has done for some time. McKinsey’s 2018 study of digital manufacturing found that the gap between piloting and full roll-out was wider than the gap between recognising a technology as relevant and piloting it, with fewer than a third of surveyed companies having moved from pilot to scale. The bottleneck sat after the experiment, not before it.

A more recent reading of an adjacent subject shows the same shape. McKinsey’s 2026 global survey on artificial intelligence reports that the share of organisations scaling AI across the enterprise rose to 44 per cent from 38 per cent a year earlier, while the share reporting any enterprise-level EBIT contribution stayed flat at 37 per cent. That survey covers internal technology deployment rather than collaboration with external partners, so it is cited here for the pattern only: deployment activity rises, recorded organisational effect does not follow it.

What actually failed: predictability at the handover

The Innovation Flow framework describes four conditions that determine whether an idea moves: legitimacy, predictability, connectivity and innovation memory. At the integration stage, the second study places implementation pathways, ownership, budget responsibility, resource allocation and decision-making processes under predictability. Innovation memory answers a different question at this stage: why the same failure repeats across projects year after year.

So the condition that fails when the portfolio grows and production does not is predictability, and it fails through four mechanisms that can be observed separately.

Ownership transfer gap

Pilots are frequently initiated by innovation or transformation teams, while operational implementation depends on business units responsible for infrastructure, execution, budgets and long-term support. The pilot therefore has an owner for its duration and no owner for its afterlife. The handover is treated as a communication event rather than as a transfer of responsibility, and the receiving unit has usually not agreed to receive anything.

“Projects lose momentum when they move from innovation teams to business units.”
Head of R&D · Energy · Norway

Budget ownership

A pilot may be technically validated and still fail to move forward if no business unit, profit-and-loss owner or operational sponsor is prepared to fund scaling, allocate people or absorb implementation risk. The money that paid for the pilot usually came from an innovation budget built for experiments, with no provision for run costs, licences, integration work or second-line support. The receiving unit is asked to take on a recurring cost that was never in its plan, in exchange for a benefit calculated by someone else.

Distributed decision-making

Decision-making is frequently distributed across multiple units with different priorities, which slows continuation and weakens accountability. The information about the pilot travels perfectly well. The authority to act on it does not travel with it.

“Information moves quite well, but decisions are still made within units. That’s where things slow down.”
Digital Innovation Leader · Telecommunications · Finland

Encouragement without commitment

The fourth mechanism is the most expensive, because it produces no visible refusal. One interviewee described the question that has to be asked and usually is not.

“Do you understand how this future could look? And are you willing to commit to it?”
Director of Growth & Development · Energy Technology · Netherlands

Without that step, as the same interviewee put it, initiatives receive encouragement of the kind that sounds like approval and commits nobody, while no one has any real intention of putting resources behind them. A project in that state is worse than a rejected one. It keeps a slot in the portfolio, consumes attention from the innovation team, and occupies the external partner’s roadmap, all without a decision having been made.

The second layer: why it repeats

Predictability explains why a given pilot stops. Innovation memory explains why the organisation is in the same position next year. Implementation learning at this stage is distributed across teams, countries and user groups: some units adapt quickly, others meet resistance or different operational constraints. Unless that learning is captured and carried across units, organisations restart integration learning each time rather than accumulating capability for embedding.

The academic reading of this is older than the vocabulary. James March’s 1991 account of exploration and exploitation describes the structural tension precisely: the returns from exploitation are closer in time, more certain and easier to attribute, so organisations systematically under-invest in the transfer from one mode to the other. Cohen and Levinthal’s 1990 work on absorptive capacity adds the reason that the receiving unit often cannot evaluate what it is being given. Steven Kerr’s 1975 paper on rewarding one behaviour while hoping for another covers the rest: an organisation that counts pilots started is rewarding the act of starting, and it will get more starts.

Table 1. Requirements for effective integration-stage collaboration by Innovation Flow condition. Source: Bridgium, From Discovery to Practice (2026), Table 5, section 7.5.

How to check this in your own organisation

Five questions are enough, and each can be answered in a single conversation. None of them requires a survey or an external review.

• First: for the pilot that finished most recently, who is named as the operational owner, and in which document is the name written? An owner who exists only in a conversation is not an owner.
• Second: which budget line pays for the first twelve months of running that solution, and whose profit and loss does that line sit in? If the answer is still the innovation budget, the handover has not happened.
• Third: what date was set for the continuation decision, and who has to be in the room for it to count? A decision with no date is a queue position.
• Fourth: what was the recorded outcome of the pilot before last, and where would a colleague who joins in six months find it? If the answer is a person rather than a place, the organisation has no innovation memory, it has institutional recall that leaves when that person does.
• Fifth and this is the diagnostic one: what was the last pilot the organisation formally stopped? An organisation that has never formally stopped a pilot does not have a continuation process at all. It has an accumulation of open items, and the cost of each is invisible because nothing was ever decided.

The Nordic dimension

Both the advantage and the risk in Nordic organisations come out of the same sentence from a Finnish interviewee, quoted above. The first half is the advantage. Information moves quite well: flat structures and high internal transparency mean findings from an innovation team reach operational units without formal escalation, and a pilot result is generally known well beyond the team that ran it. Compared with organisations where visibility itself has to be fought for, the Nordic starting position at this stage is good.

The second half is the risk. Decisions are still made within units, and consensus norms make it comparatively easy to obtain agreement in a room and comparatively hard to obtain a named owner outside it. The fourth mechanism above, encouragement without commitment, travels particularly well in a culture where a direct refusal feels impolite and a warm response costs nothing. The result is a pilot that everybody supports and that no unit has scheduled.

Table 2. Observable symptoms and the checks that distinguish an entry-stage problem from an integration-stage one. Source: Bridgium analysis based on From Discovery to Practice (2026), section 7.

The structural response

Four responses follow from the mechanisms above. All of them are changes to when a decision is made rather than to how much is spent.

1. Name the operational owner and the budget line before the pilot starts. Successful integration depends on clear ownership assignment, and the assignment is far cheaper to obtain as an entry condition than as an exit negotiation. Before a pilot is approved, one business unit should have agreed in writing that if the pilot meets its criteria, that unit takes the solution, the run cost and the support obligation. A unit that will not sign that before the work starts will not sign it afterwards, and knowing this in advance saves several months of everyone’s time.
2. Publish the continuation criteria and the decision date with the pilot brief. Transparent implementation criteria and visible decision-making pathways let innovation teams, operational units and external partners align around the same execution goal. Transparency here means visibility of the criteria and of who decides. It does not mean disclosing strategy, and the distinction matters because the second is usually what gets refused when the first is requested.
3. Convene a cross-functional alignment group for the transition window. Procurement, legal, IT, compliance and the receiving operational unit should meet the solution before the pilot ends rather than after it. This turns the structural gates described above from surprises into scheduled work, and it gives the external partner time to produce the reliability evidence that will be demanded of it.
4. Record the implementation outcome where the next team will look for it. Regular feedback loops and mechanisms for capturing implementation learning are what allow an organisation to build innovation memory instead of restarting adoption each time. The requirement is modest: a short standing record of what was piloted, what the continuation decision was, who made it and what happened next, kept in a location that does not depend on any individual remaining in post.

None of these four requires additional budget, and none of them changes the pilot itself. They change the point at which the organisation finds out whether anyone intends to use the result.

Conclusion

A growing pilot portfolio is often read as evidence of innovation capability. It is more accurate to read it as a measure of how well the organisation starts things. Starting is the part that the innovation function controls, and it is therefore the part that gets measured and rewarded. The transfer of responsibility to an operational owner is the part that no single function controls, and it is therefore the part that fails without anyone recording a refusal, while every individual metric stays healthy.

What changes when ownership, criteria, cross-functional timing and a written record are in place is not the success rate of pilots. Some solutions will still be declined, and a portion of them should be. What changes is that the decline happens on a date, with a name attached, early enough for the innovation team to redirect its attention and for the external partner to plan. An organisation that stops four pilots deliberately in a year is in better condition than one that stops none and carries nineteen open items into the next planning cycle.

The next quarterly review has a better question available to it than the count of active pilots. For the pilot your organisation finished most recently, can you name the person whose budget pays for its second year?

The findings above come from From Discovery to Practice: How Corporate–Startup Collaboration Becomes Usable, a Bridgium study based on 48 interviews with corporate innovation leaders and startup founders across Northern and Central Europe. Section 7 covers the integration stage in full, including the requirements table reproduced above. The report is available here:
https://bridgium-research.eu/startup-report-2026/

References

1. Bridgium, From Discovery to Practice: How Corporate–Startup Collaboration Becomes Usable (2026), section 7
2. Bridgium, How Innovation Happens: Insights from Leading Enterprises in Times of Change (2026)
3. McKinsey & Company, How digital manufacturing can escape ‘pilot purgatory’ (2018)
4. McKinsey & Company, The state of AI in 2026: On the road to ROI (2026)
5. McKinsey & Company, The state of AI in 2025: Agents, innovation, and transformation (2025)
6. March, J. G. (1991). Exploration and Exploitation in Organizational Learning. Organization Science, 2(1)
7. Cohen, W. M. & Levinthal, D. A. (1990). Absorptive Capacity: A New Perspective on Learning and Innovation. Administrative Science Quarterly, 35(1)
8. Kerr, S. (1975). On the Folly of Rewarding A, While Hoping for B. Academy of Management Journal, 18(4)
9. Weick, K. E. (1995). Sensemaking in Organizations. Sage
10. Nonaka, I. & Takeuchi, H. (1995). The Knowledge-Creating Company. Oxford University Press
11. Berger, P. L. & Luckmann, T. (1966). The Social Construction of Reality. Penguin
12. Knowledge loss and employee turnover, Part I. The Learning Organization, 30(2)
13. Staw, B. M. (1976). Knee-Deep in the Big Muddy: A Study of Escalating Commitment to a Chosen Course of Action. Organizational Behavior and Human Performance, 16(1). Print edition