AI tools can reduce the time and expense required to create software, prototypes and early product experiments. Yet many organisations still struggle to turn that lower build cost into faster learning. The constraint is often no longer the ability to make something; it is the time required to secure approvals, interpret evidence and make a decision.
Dan Toma frames this issue through Innovation Accounting: an approach to managing innovation by measuring learning and progress rather than relying only on conventional financial measures. His argument is especially relevant in the AI era. When teams can test more ideas at lower cost, slow governance can become a larger share of the total cost of innovation.
When cheaper experimentation exposes expensive bureaucracy
The central paradox is straightforward. Generative AI and related tools may help teams draft code, analyse information, create content and develop early concepts more quickly. But an experiment does not create value simply because it is built. Its value lies in what the organisation learns from it and what it chooses to do next.
If a team must wait weeks for a review meeting, prepare extensive approval materials or seek sign-off from multiple committees before acting on a clear result, the saving made during development can be lost in the decision process. In this context, bureaucracy is not merely an inconvenience. It is a source of delay that can raise the cost of learning.
That is the new cost of failure: not only the money spent on an unsuccessful initiative, but also the organisational time consumed before a weak assumption is recognised, stopped or revised. A small experiment that produces a negative result quickly may be relatively inexpensive. A project that continues because nobody has a timely mechanism to interpret and act on the evidence may become much more costly.
Innovation Accounting: measure learning, not activity
Innovation Accounting offers leaders a way to distinguish activity from progress. Rather than treating a completed prototype, a presentation or a large volume of features as proof that an innovation initiative is advancing, the approach asks what uncertainty has been reduced and what evidence has been produced.
For an AI-enabled initiative, this can mean identifying the assumptions that matter most before investing heavily. These may include whether a customer has a problem worth solving, whether a proposed solution is useful, whether a team can reach a defined audience, or whether the operating model can support the idea responsibly.
The practical question is then not simply, “Did the team deliver?” It is, “What did the team learn, how reliable is that learning, and what decision should follow?” This reframes failure. A test that invalidates an assumption early can be useful because it prevents further investment in an idea that lacks sufficient support.
What leaders should measure in an AI-driven organisation
Conventional measures remain important, particularly for established operations. However, they may be incomplete for work that is still searching for a viable problem, customer proposition or business model. Leaders can supplement delivery and budget reporting with measures that reveal the speed and quality of organisational learning.
- Time to decision: How long does it take to move from an experiment’s result to a documented decision?
- Decision latency: Where do approvals, reviews and hand-offs accumulate after evidence is available?
- Assumptions tested: Which high-risk assumptions have been tested, and which remain unexamined?
- Learning quality: Did an experiment produce evidence strong enough to support a decision, or only output and activity?
- Cost of delay: What resources continue to be consumed while a team waits for a decision or permission to proceed?
These measures do not imply that organisations should remove governance. AI initiatives can involve material concerns around risk, data, customers and accountability. The challenge is to design governance that is proportionate to the stage of uncertainty and clear about who can make which decisions.
Reducing governance friction without removing accountability
A useful starting point is to separate small learning experiments from commitments that are difficult to reverse. A limited test, with defined boundaries and clear success or stop criteria, may not need the same approval route as a large-scale deployment. Giving teams decision rights within agreed limits can reduce unnecessary escalation while preserving oversight for higher-risk choices.
Leaders can also make evidence requirements explicit. Before an experiment begins, agree on the assumption being tested, the evidence that would change a decision and the person accountable for acting on the result. This helps avoid a common problem: teams generate information, but no one is obliged to decide what it means.
Finally, review forums should be designed to make choices, not merely receive updates. If a committee exists to assess innovation work, it should have a defined mandate, a regular cadence and the authority to approve, redirect, pause or stop initiatives. Otherwise, reporting can become another layer of delay.
Why Dan Toma’s innovation speaker perspective matters
For CEOs, CFOs, product leaders and transformation teams, Dan Toma’s Innovation Accounting lens raises a timely management question: if AI makes experimentation cheaper, has the organisation also made learning and decision-making faster? It moves the conversation beyond AI adoption rates and feature delivery towards the operating conditions required to use experimentation responsibly.
This topic is suited to leadership meetings, innovation forums and corporate events focused on AI strategy, digital transformation, portfolio management and organisational agility. Event audiences can use it to examine their own approval systems, decision rights and innovation metrics rather than treating AI as a technology issue alone.
To enquire about Dan Toma for a conference, leadership meeting or corporate event, contact Speakers Connect at info@speakersconnect.com.