You Have 40 AI Pilots. Which Ones Deserve the Next Riyal?

At GPMF in Riyadh last month, AI came up in almost every session I went to. What I heard much less about was how organizations decide which pilots actually get funded once twenty or thirty of them are competing for the same budget.

I spent two days at the Global Project Management Forum, and AI was part of nearly every talk I sat in and most of the conversations in the hallways.

The question I kept thinking about was a practical one. Of all the AI pilots an organization is running, which ones get funded to scale next year, and who decides?

In most of the organizations I work with in Saudi Arabia and the UAE, there isn't a clear answer to that yet.

Picture a large enterprise in Saudi Arabia with 30 to 40 AI pilots running. A few started because a vendor offered a free proof of concept. Others came from a business unit that moved quickly, or from the board or a ministry wanting to see progress on AI.

Once you count internal time, each pilot costs around SAR 1.5 million, so the total comes to SAR 45 to 60 million a year.

Gartner surveyed more than 3,100 CIOs in 2024 and found that only 48% of digital initiatives meet or exceed their business outcome targets.¹ I don't see a reason to expect AI pilots to do better. At that rate, roughly half of a pilot budget that size doesn't deliver the outcome it was meant to.

We call this the Wrong Work Funded gap, and across the portfolios we see it's worth about 3.5% of EBITDA.² AI is making it bigger, mostly because pilots are cheap to start and hard to stop.

In my experience there are three reasons a list of pilots never turns into a portfolio.

Too many ways in. Use cases come in through IT, innovation teams, business units and vendors, each with its own template. They're never compared side by side for the same money, so the sponsor with the best connections usually gets funded.

Reporting on activity. Pilots get reported by what the team did, like models trained, users onboarded and demos run. That shows the team was busy, but it doesn't tell you whether costs, cycle times or revenue actually changed.

Nobody owns it after sign-off. In the same Gartner survey, organizations where business executives co-own delivery with the CIO hit their outcome targets 71% of the time.¹ In most pilot programs I see, the business sponsor signs the request and then steps back. When the pilot ends, nobody on the business side is responsible for getting people to use it or for checking whether the benefits showed up.

"AI initiative" covers a lot of very different things. Before anything gets funded, we find it helps to sort them into three groups.

Each group needs its own business case, owner and test. If they all sit on one list, you end up comparing a copilot rollout with a new revenue product on the same slide, and that comparison doesn't help anyone decide.

The organizations I see doing this well fund AI the same way they fund everything else. The money comes in stages, and a pilot only moves to the next stage when it has the evidence for it.

Two things make this work in practice. First, agree the stop criteria at intake, before anyone gets attached to the pilot. Second, review all the AI work together every quarter, so each use case is compared with the others and with the rest of the investment portfolio.

In Conductor, every initiative, AI or not, comes in through the same intake, is linked to a strategic objective and moves through stages tied to funding. Harmony AI flags duplicates and new risks while the portfolio runs, and outcome tracking starts the day a use case is approved.

At EtihadWE, bringing more than 50 initiatives into one orchestrated portfolio gave leadership 100% real-time visibility into what was funded, why, and how it was performing.³ A major utility in France has built schedule deviation tracking and delay classification into its stage gates in Conductor, so each slip is classified when it happens and leadership can see which projects are drifting and why.

If your board asked tomorrow which AI pilots will scale next quarter and why, how long would it take you to answer?

If it would help, we can sit down with your team and run two or three of your live use cases through the three groups and the four stages.

Sources

¹ Gartner, "Only 48% of Digital Initiatives Meet or Exceed Their Business Outcome Targets," October 2024.

² Sensei Labs, Four Gaps analysis of client portfolios.

³ Sensei Labs client data, EtihadWE. Pilot count and cost are illustrative.