Three months in, the update lands with confidence. The AI programme is gaining traction, the team is finding their feet with the tools, the vendor relationship is developing well. You walk away without a financial number in sight, unsure whether you have just received an honest assessment or a well-dressed stall.
Early in any AI programme, hard results are not yet there, and they shouldn’t be. Meaningful ROI typically takes 12 to 24 months to show in the margin. The honest answer to “where is the return?” at this stage is “not here yet.” A founder who treats that as failure trains the delegate to never say it again.
But not every no-results-yet update is equal. Some are credible, some are theatre. This post helps you tell the difference.
What does “no results yet” actually mean?
“No results yet” describes a real and expected phase in any serious AI programme. In the first three months, the work is building the conditions for results, not generating them. Capability gets built, data gets cleaned, risks get managed. The gap between adopting a tool and seeing it affect the margin is real, documented, and often longer than many boards anticipate.
BCG’s 2025 analysis found that AI usage has risen sharply while financial impact has not followed. The OECD’s research on AI adoption in smaller businesses notes that owner-managed firms often start from a lower data infrastructure baseline than larger organisations, extending the build period further.
The build phase has real deliverables. A team that can use the tools reliably. Data clean enough for the specific task. Documented workflows someone other than the delegate can follow. A governance model that covers the obvious risks. When those are in place, results tend to follow. When they are not, waiting changes nothing.
Why does your reaction to this update matter?
The delegate giving you a no-results-yet update is watching how you receive it. If you respond with visible frustration, demand a faster timeline, or pull in a more optimistic external source to challenge their read, they learn a lesson. The next update gets reframed. Small wins get amplified. Risks that are still open get softened. The information that would help you make decisions starts getting managed.
Research from Korn Ferry on AI readiness in organisations shows that delegates are often assigned for their operational delivery skills, not their AI expertise, and they are carrying real professional exposure as a result. They know the pilot-to-scale success rate is low. They are reading the founder’s signals carefully. Punishing honest no-results-yet updates pushes a capable person towards managing perceptions rather than managing the programme.
Spencer Stuart’s research on how founders handle the AI mandate in their businesses documents the pattern clearly. The founder who disengages until results fail, then re-engages with unrealistic correction demands, creates the conditions for polished updates rather than useful ones. The delegate carries the risk while the founder holds the resources. That combination, when honesty is also punished, produces managed communication rather than honest communication.
Research from HRExecutive on AI-related cultural signals reinforces the same point. When leadership responds poorly to honest reporting about AI progress, the information environment adjusts.
What should a credible update actually contain?
A credible no-results-yet update has three ingredients. Named risks managed, capability built, and leading indicators moving. Each one does specific work. Risks managed shows the delegate is thinking ahead. Capability built shows the infrastructure for results is actually there. Leading indicators show direction without overstating arrival. A delegate who can give you all three with specific examples from your own firm is running a real programme.
Named risks managed means specific risks were identified and addressed, not a general statement that risks were considered. A data quality issue that would have broken a pilot was caught and resolved. A compliance question about how a particular AI tool handled client data was answered before the tool went into production. A workflow dependency nobody had mapped was found and documented. The specificity is what separates a managed risk from a waived one.
Capability built means the team can use the tools reliably, the data is clean enough for the specific task, and someone other than the delegate can run this if they leave. Logixguru’s 90-day AI readiness framework identifies five dimensions covering data maturity, team enablement, technology infrastructure, strategic alignment, and governance. A credible update gives you an honest read across at least three of these.
Leading indicators predict commercial value without yet proving it. Propeller’s dual-ROI framework distinguishes trending ROI, the early indicators that a programme is working, from realised ROI, the financial outcomes that follow. Leading indicators include adoption rates, time saved on specific tasks, reductions in error rates, and accuracy scores on outputs. A delegate who can show task completion time dropping 40% on a specific workflow is showing you something real, even before that saving reaches the margin.
What are the red flags in a hollow update?
A hollow no-results-yet update substitutes optimism for evidence. Three patterns show up reliably. The first is pure enthusiasm with no underlying indicators at all; things feel positive but there are no numbers, even soft ones. The second is heavy reliance on external sources rather than the firm’s own read. The third is a moving definition of success, where what was due by month three is now a six-month target, with no explanation of why.
A credible three-month update will not have financial results. It will have numbers. Adoption percentages, time saved on specific tasks, accuracy comparisons before and after, error rates on a particular workflow. These are the things a delegate who is genuinely running a programme has to hand. An enthusiastic update with none of them has no foundation you can evaluate.
Citing BCG data or a vendor case study is not the same as reporting what is happening in your firm. When an update is built almost entirely on what others have observed elsewhere, it often signals that the delegate does not yet have an independent read on their own programme. That is sometimes expected in the very early weeks. When it persists past month two, it suggests the delegate is covering a gap between what they know and what they are being asked to report.
A well-run programme sets specific milestones at the outset, covering what capability will be built by when and which indicators will be tracked. If those milestones have shifted without explanation, ask what changed and why the original target was wrong. A delayed milestone is not always a red flag. A delayed milestone the delegate has not named as a delay usually is.
How should you receive an honest update?
When the update is credible, the most useful thing you can do is say so explicitly. Name the three elements you heard. Saying something like “you’ve covered risks managed, capability built, and indicators moving, and that is what I need from this stage” costs you nothing and does significant work. It tells the delegate that honest reporting is what earns your confidence, not optimistic framing.
The follow-up question matters too. Asking “what is the leading indicator you are watching most closely right now?” or “what is the biggest risk still open?” rewards honest programme thinking rather than performance. Those questions differ from asking “when will we see results?” which tends to produce a date that protects the relationship rather than one that reflects the actual position.
Hold back external benchmarks in these conversations. If the board has raised the question of returns, or a peer company has shared early AI numbers, keep that out of the update meeting. Introducing external comparisons pushes the delegate towards matching a benchmark rather than reporting their own read.
Addepar’s framework for executive questions during AI adoption is worth reading here. The useful questions go upstream of the result. What problem are we solving? What does the programme look like when it is working? What would cause it to stall? These are the questions that help you think clearly about the programme’s actual state, rather than questions that create pressure around it.
EY’s research on AI governance and board engagement identifies information quality between programme sponsors and programme leads as a distinguishing factor in which programmes eventually deliver. A founder who has made honest reporting feel safe has a better read on where the programme actually is when the real decisions arrive.
