The easier the research gets, the less your team fights for it. Let them shape the decision, and they’ll stand behind it.

At Innov8rs Manchester, Vincent Atallah, co-founder and president of the AI innovation platform Aucctus, asked a room of innovation leaders three questions. Most raised a hand to say they had read an AI output and forgotten it by the next meeting. Most also said they felt buried by how much AI can produce. Only one person said they would feel confident taking an AI output straight to leadership and defending it.

A study of 119 innovation teams by Innov8rs and Aucctus found the same gap: 97% use AI, yet only 8% let it inform decisions to kill, continue or invest in an idea. Vincent’s explanation for the gap: teams don’t fight for conclusions they didn’t reach. AI now does most of the work, so teams have little to defend. His answer is friction: small, deliberate steps where people do part of the work themselves, so they own the result.

He suggests adding those steps at four points where an idea is most at risk: when the team checks its assumptions, before the research starts, at the gate meeting and in early talks with senior leaders. A short check at the end shows whether the friction is working.

Fixing the wrong problem three times

Vincent’s company builds AI that does innovation research, including the list of risky assumptions behind each idea. For two years, his team worked on removing friction. First, Aucctus turned four months of research into eight minutes. When users said the AI hallucinated, the team made every claim cited and sourced. When users said the AI missed how their business works, the team built a way to load each company’s context. “Gosh, we’ve done enough, haven’t we?” Vincent remembers thinking.

Even so, teams rarely used the outputs to make decisions. Users read the assumptions and moved on. They left them out of meetings, and when they tested one, they tested it once. Aucctus kept improving the assumptions until they were close to flawless, and users still ignored them.

The research was good, so the team looked at how ideas actually move forward inside a company. In a brainstorm, the idea that gets picked is often the one with a motivated person pushing for it.

❝

“Conviction, whether we like it or not, is something that moves mountains in the innovation process,” Vincent says.

Users had done little of the work behind the AI’s assumptions, so they felt little push to fight for them. That led the team to a new question: “Are we missing friction?”

David Bland, co-author of Testing Business Ideas, helped Aucctus answer that question and redesign the assumptions feature. “Nobody defends work that they didn’t do,” he told the team. The line was hard to hear for a company whose product promises to do the work for its users.

From then on, Aucctus made friction a design rule. For every new feature, the product team asks which friction to remove and which to add back on purpose.

Make people approve every assumption before they move on

Aucctus changed the assumptions step in two ways:

  • Each company sets its own bar for weak and strong evidence across desirability, viability and feasibility.

  • Users approve or reword each assumption before they can move to the next one.

Teams can apply the same rule without software: no assumption goes into testing until the person who owns it has approved or rewritten it.

These small changes make users feel deeply connected to their assumptions. One customer heard a colleague suggest a change to an assumption she had written. “Don’t touch my assumption,” she said. “I spent so much time on that assumption.” That level of ownership is the goal, but only up to a point: a team should stay attached to an assumption only as long as the evidence supports it.

Aucctus’s own user data shows three changes since the redesign:

  • Users who go through these steps are five times more likely to recall the top three risks in their own concept.

  • Users edit 85% of assumptions before approving them. Even a one- or two-word edit can change what an assumption means and which test the team runs.

  • Average sessions grow from 11 to 43 minutes, because teams spend the extra time debating what decides the concept’s success. Aucctus reports stronger results later in the funnel.

Make the team commit to the result that kills the idea

Teams that set the bar after the data arrives tend to convince themselves they were right. That habit is confirmation bias; a team that decides afterwards will call the result “directionally positive.” Anyone trained in testing business ideas knows the rule, yet teams skip it all the time.

The added friction is one sentence, written before the prompt, survey or customer interviews: if this happens, we’re right; if that happens, we’re wrong.

This means agreeing on what counts as strong evidence. Every company draws that line differently. Schreiber Foods, a large dairy company, considers a concept only if it can reach a set gross margin, and its executives ask about that threshold at every presentation. Rules like this often go unspoken inside a company. The approach asks teams to set both levels, a weak signal and strong proof, before any testing starts. Take a software product that dairy farmers use to log what they feed their cows, so their buyers can report carbon cuts. Before testing it, the team sets two bars:

  • Weak signal, what farmers say: In interviews, 8 of 10 farmers say they would log their feed data if paid. Too thin to commit money.

  • Strong proof, what farmers do: In a small hands-on pilot, 12 of 15 farmers actually log their feed data every week for a month. Enough to commit.

Make the person who built the case defend it at the gate

Many teams send an executive or manager to update leadership on concepts. That handoff can be a mistake, because executives judge an idea partly by the conviction of the person presenting it. That’s why the people who did the work with customers should present the concept themselves.

The rule of thumb is “whoever presents, writes.” The person defending the concept at the gate writes the risks by hand, and AI then challenges those risks.

Pre-reads create a second problem in the same meeting. AI makes it easy to send executives a polished summary in advance, so they arrive with their own biases. “They come into the meeting room having already made a decision,” Vincent says. The debate ends before the meeting starts, and the room loses the executives’ first reactions to the key points. Skipping the pre-read keeps both the debate and those reactions in the room.

Bring the senior leaders who could stop the idea into the friction early

Teams often keep senior leaders out of the process because they want to arrive with a polished, finished pitch.

If a senior leader is likely to be the one who shuts the idea down, Vincent’s advice is to bring that leader in early. The leader then debates the idea with the team and helps decide along the way, instead of seeing it for the first time as a finished pitch.

❝

“The friction that you create with those senior leaders ends up strengthening the idea and the championship you might receive down funnel,” Vincent says.

Open debate also helps the team’s standing, because senior leaders value seeing a team that disagrees openly and takes part in the process.

Diagnostic: the friction audit.

Run this before a concept goes to its next gate.

  • Three-answer test: Ask three people, separately, to name the concept’s biggest risk. A team that owns the case names the same risk. Three different answers, in Vincent’s words, means “you don’t have a team; you have a distribution list.”

  • Zero edits: Open the version history on the last AI document that went to a gate. Zero edits usually means the team skipped reading it, or cared too little to challenge it.

  • Recite vs. defend: Ask your sponsor why the recommendation is right. A champion makes the case in their own words; a narrator just repeats the summary.

A concept that fails any of these checks has a friction problem.

Three fixes can start at the next test and the next gate: set the kill result before the research starts, have the person who wrote the assumptions present them, and keep the summary out of executives’ inboxes until the meeting.

Reply

Avatar

or to participate

Keep Reading