Why R&D execution is becoming a human discipline again
Why R&D execution is becoming a human discipline again
For a moment, let’s take a mini break from Artificial Intelligence (AI). I know how that sounds given how obvious it is as a focus area, but almost every Research & Development (R&D) leader I speak with is still facing a human opportunity, and frankly I don’t think we are talking about it enough.
The technology available to R&D and innovation teams is improving quickly. AI can help generate hypotheses, scan technical literature, accelerate simulation, support experiment design, and reduce some of the manual challenges that has always slowed research. Lab automation is becoming more practical. External innovation ecosystems are developing and are more accessible. Data is richer, even if it remains frustratingly fragmented.
However, inside many large organizations, innovation outcomes still move at the speed of the meeting cycle. In some cases, worse.
Reach out to discuss this topic in depth.
That is the mismatch I want to talk about in this blog. The evidence base for a program can now shift quarterly, monthly, or in some domains weekly. The funding and governance systems wrapped around that program still move annually. The operating model for R&D is changing, but many organizations continue to treat execution as a project management problem. Set the annual budget, approve the portfolio, track milestones (hopefully), review progress, ask for evidence, then continue, pause, or cut. That model certainly provides the appearance of discipline. It also hides a healthy amount of avoidance.
Most large organizations can already produce a plentiful number of ideas, more data, and more technical options. What is much less certain is whether the organization can change its mind quickly enough when the signals change. To me, this is where the real human work rests, and it is also where the system is most fragile.
It is a reality that whether you want it or not, by proxy with the accessibility of AI you have a faster innovation system, but that creates more moments where judgment is exposed. Leaders have to decide whether an early signal, of any kind, not just technical, is meaningful or not. Teams have to say when a hypothesis has weakened, even after spending two years defending it. Finance has to understand why optionality has value before a conventional business case can be built. Senior sponsors have to let go of programs that once made sense but now consume capital, attention, and talent out of proportion to their future relevance. And the timetable for all of this is condensing rapidly.
We do love our dashboards. But none of this is solved by a better dashboard. In fact, more data can make the problem worse if the organization lacks the confidence to interpret it. I have seen portfolios where the reporting is extensive, the governance is maturing on paper, and the actual decision quality is poor. Everyone can explain what is happening, but far fewer people are willing to say what should stop.
This is why the old R&D incumbency trap matters so much. Existing programs are easier to defend than new bets are to launch. They have teams, history, sunk cost, internal advocates, supplier relationships, and enough progress to look credible in a review. A new opportunity, by comparison, often arrives as an inconvenient signal. It does not have a full business case. It may not fit the current capability map. It may threaten a program that already has executive sponsorship. Honestly, it threatens our emotional attachments. We are, after all, human. If left unchecked, the portfolio slowly becomes eroded.
That sounds harsh, but many R&D leaders will recognize this cadence. The bold bet from five years ago becomes a protected line item when a program that should be challenged becomes politically expensive to question. The team that has learned the most about why something may not work is still asked to present why it might. And worst of all, I have watched leaders ask AI to support or argue a point without realizing the AI is itself sycophantic. It will happily polish a conviction that deserved to be challenged.
Modern R&D execution needs a more honest relationship with learning. If AI and other digital tools compress parts of the research cycle, then governance has to become responsive to learning cycles. Funding cannot remain trapped in stubborn annual systems while the evidence changes on an opposing cadence. Portfolio reviews cannot reward activity when what leaders really need is sharper conviction. For leaders to continue in this era, they must be spending more time asking whether the original belief still deserves capital.
That is a different kind of conversation, and I understand why some may find it irritating. Why should a team keep justifying a concept that has already been approved?
Because this is about trust between technical teams and leadership. It requires permission to identify uncomfortable evidence before it has been crafted into a nice slide. It requires Chief Financial Officers (CFOs) to see innovation funding as a portfolio of changing probabilities rather than a queue of projects asking for protection. It requires leaders to stop confusing persistence with discipline.
Handled well, none of this makes R&D less human. If anything, it makes human judgment more explicit and puts it back where it belongs, as the primary decision factor.
In practice, that looks like a few specific disciplines. Separate core improvement from exploratory work and fund them differently. Hold back strategic reserve capital so you can act when new signals emerge. Define kill criteria before emotional attachment hardens. And ask every major program, repeatedly, what evidence would increase conviction and what evidence would reduce it.
I view that last part as mattering most. If no evidence could change the decision, the governing has already stopped. What remains is protection, and protected programs should be first in line for sunset.
This is also where the conversation with the CFO has to develop. R&D leaders keep asking finance for patience in some generic sense, and I think that is the wrong request. The better move is to show how a more dynamic execution model reduces wasted capital by surfacing risk earlier. A rolling, signal-driven portfolio gives the organization more chances to redirect spend before large commitments become politically and operationally difficult to unwind.
To me, the future R&D leader looks less like a sponsor of projects and more like an architect of conviction. They shape where belief is allowed to form, where it has to be tested, where it earns more capital, and where it is retired with discipline. They understand the science and the economics. But they also understand the emotional life of the organization: pride, fear, attachment, status, fatigue, and the quiet pressure to keep funding what already exists.
That is the humanity in this shift. Some will hear it as soft leadership. I think it is simply more honest leadership.
The next phase of R&D execution will have more AI in it, of course. More automation, more simulation, more data, more external signal. But the real advantage will go to the organizations that can turn faster learning into better decisions without losing the people inside the system.
If you enjoyed this blog, check out, Rethinking how research supports R&D decision-making – Everest Group Research Portal, which delves deeper into another topic relating to R&D.
If you’d like to continue this discussion, please contact Richard Sear ([email protected]).