100 days show whether the AI bet pays off

The first 100 days after closing are the period of highest organizational attention a portfolio company will ever have. Management expects change, the board is watching closely, and decisions can be made that would meet resistance three months later once daily routine has settled back in. Anyone who fails to translate an AI thesis from the investment memo into a running system in this window loses more than time. The window in which change was easy closes.

For the investment committee, the 100-day plan is also the first hard evidence of whether the AI thesis assumed in the deal actually holds up. Value creation plans often contain a paragraph on AI-driven efficiency. Whether that paragraph turns into a number is decided in this first phase, not in year three.

That holds regardless of how technically mature the target already is. A company with a modern system landscape can get further in 100 days than one with an accumulated legacy architecture, but both benefit from delivering early proof instead of an early announcement. The difference lies in the lever chosen, not in whether one is chosen at all.

The most common mistake is starting with the most ambitious initiative: full automation of a core process, a company-wide AI platform, a new data architecture for every use case at once. It sounds compelling in a board meeting. In practice it ties up the first weeks in architecture debates while no visible result appears. After 100 days, what remains is a concept slide, not a production system.

Big initiatives need trust that a new management team does not yet have, not from the board and not from its own workforce. That trust comes from a visible, working result, not from a persuasive presentation.

The second effect of the big initiative is more subtle: it ties up the best people in the company in debates about the right architecture while day-to-day work carries on unchanged. After 100 days, nothing has changed in the cost structure or the error rate, only the slide deck has gotten more detailed.

The first two weeks belong to a tight stocktaking exercise: which data is actually accessible for the AI levers assumed in the deal? Which systems would an agent or automation workflow need to reach, and how clean are their interfaces? And, often underestimated, where are individual employees already using AI tools on their own initiative, without management's knowledge? This shadow usage reliably shows where the real need lies.

If the tech due diligence already produced an architecture picture, that is the starting point. If not, this stocktaking is the condensed version of it, tight enough to fit into two weeks, honest enough to name the risks.

Part of this stocktaking should also clarify which decisions the existing management can make and which need sign-off from the board or the operating partner. Clarifying this in week one saves valuable time in week four that would otherwise be lost chasing approvals.

The stocktaking leads to a choice: a single lever from the value creation plan that runs in production by day 100. The criteria for that choice are sober: high frequency in day-to-day operations, an outcome measurable before and after, manageable integration risk, and an accountable owner in management who carries the result. A customer service agent for the most common request category usually meets that bar. A company-wide overhaul of order processing rarely does.

These weeks are execution, not concept work. That also means the lever has to be chosen so it doesn't require months of upfront work on data infrastructure. If the data is missing, that is itself the first finding of the 100-day plan. The lever shifts, not the deadline.

This phase also reveals whether the chosen approach fits the existing system landscape. A lever that connects cleanly through an existing interface is in production within four weeks. A lever that first requires a new interface almost always blows the deadline. One more reason to take integration depth seriously at the selection stage in week two, not only after the first setback in week four.

A pilot without a metric is an anecdote. The metric has to point to an EBITDA driver: cost per case, cycle time, error rate, hours freed up on the team. Vanity metrics like "number of automated tickets" work in an internal meeting but not in front of an investment committee asking about impact on the number.

The baseline matters just as much. Without a clean measurement of the state before the lever, the effect afterward can't be proven. That baseline belongs in week 1 of execution, not at the end. Otherwise it gets reconstructed after the fact and loses credibility.

For reporting to the investment committee, it also pays to draw a clear line between what is already measured and what is still a hypothesis. A status update that names both openly is more convincing than an optimistic outlook that later has to be walked back.

A pilot that ends as a pilot after 100 days has no value for the value creation plan. The last part of the window belongs to the handover into regular operations: an accountable owner in the business function, not in the project team; an escalation rule for cases the agent or workflow can't resolve; brief training for the people who will work with it going forward. Without this step, the pilot quietly disappears in the months after day 100 as soon as the project team moves to the next priority.

The operating model is also the foundation for the second lever. A team that handles the handover cleanly the first time has a repeatable template and doesn't have to start from zero on the second use case. That template is often the real, lasting value of the first 100 days, more than the individual lever itself.

A 100-day plan without a single accountable owner unravels between the day-to-day responsibilities of the business functions. In most portfolio companies, no internal role at this early stage can judge both technical feasibility and impact on the value creation plan. That's why this role often falls to a fractional CTO or an external technical partner with a mandate from the operating partner, limited to the 100 days and with a clear mandate to hand over to an internal person afterward.

That handover is not an afterthought. A 100-day plan still carried by an external person after day 100 has not created sustainability in the company, only a single result. The brief for any external partner should therefore include, from the start, a named internal counterpart who understands the solution and can develop it further after the handover.

No company-wide platform switch before a single use case has proven it works. No build-out of a large internal AI team before it's clear which skills are actually needed. And no messaging to the workforce that equates automation with headcount cuts before there is even a result. That builds resistance to the second and third lever long before they begin. It's equally unhelpful to commission external consultants for another strategy presentation while execution of the first lever is still outstanding. Every additional slide delays execution without extending the deadline.

It's just as unhelpful to treat the deadline itself as sacred. 100 days is a rhythm, not a contract. If the stocktaking in week two shows that the originally assumed lever fails for lack of data, it's better to pivot early to a second, more realistic lever than to hold onto the original plan and end up with nothing at the end of the window.