Enterprises Gave AI Agents Their Data. 54% Still Abandoned the Projects.
Access was never the problem. Proving what the agent changed was.
Most organizations blame abandoned AI agent projects on the agents. The model was not good enough, the use case was wrong, the pilot tried to do too much.
The larger problem is what the agent was measured against. A new SnapLogic survey of 1,000 IT decision-makers in the US, UK and Germany, released yesterday, found that 54% have abandoned AI projects because business value was unclear. Yet 77% say their AI agents have, or would have, full or extensive access to the relevant data and systems.
So access is not the gap. Proof is. Every agent project eventually meets one question from finance: what did this actually change? Answering it means agreeing on a baseline, a definition and a source of record for the number the agent was supposed to move.
And it gets worse as agents multiply. Each new agent arrives with its own success metric, pulled from its own copy of the data, with its own idea of what "resolved", "active" or "saved" means.
The agents that last are the ones that finish a clear task with a result anyone can see, not a demo someone has to defend. That is what OpenMap does, and it is now live: from shopping to planning a group trip, ask in your own words and your agent does the legwork. Try it free at openmap.ai.
Why Can't Anyone Prove What the AI Agent Did?
Because the number that matters was never defined once.
Picture a service team that deploys an agent to resolve support tickets. Six months in, the team reports that average handling time has dropped sharply. Finance pulls the same metric from the billing system, where reopened tickets count as new ones, and most of the improvement disappears.
Operations has a third number, from a dashboard nobody has touched since the last reorg.
The agent worked.
Nobody can agree that it did.

The Cost Isn't a Failed Agent. It's an Unprovable One.
SnapLogic found that 48% struggle, or expect to struggle, to measure ROI from agentic AI. That is how most agent projects end: not with a visible failure, but with a budget review nobody can win.
Before anyone can say whether an agent delivered value, your team usually has to:
Agree which system holds the official version of the metric
Reconcile duplicate customer, ticket or order records that inflate or hide the result
Trace how the metric was calculated before and after the agent went live
Find who owns the definition, and whether it changed mid-pilot
Show that the data the agent acted on is the same data the result was measured on
None of these tasks are particularly difficult. They are slow, manual, and usually done after the project is already under review, when the original context is gone.
SnapLogic CTO Jeremiah Stone put the fix plainly: "Enterprises need to connect AI to the right data, applications, and workflows." The right data is not all the data. It is data everyone has agreed means the same thing.

How to Make AI Agent Value Provable in 30 Days?
You do not need to redesign your data platform to prove what your agents are worth.
Within the first month, focus on the three to five metrics each live agent is meant to move, and the entities behind them: customers, tickets, orders and accounts.
Start by discovering the metadata you already have, mapping where each metric comes from, which systems feed it, and which reports, models and agents depend on it downstream. Then resolve duplicate records so every entity is counted once, and give each definition a named owner.
Next, expose this context through a single knowledge layer that your agents, your analysts and your finance team all read from, so the baseline and the result come from the same place.
DataManagement.AI accelerates this process by automatically discovering metadata, mapping lineage across the modern data stack, identifying downstream dependencies, and creating a centralized knowledge layer that stays continuously updated as your environment evolves. It resolves duplicate and conflicting records, and ownership and policy travel with every entity, so the number your agent moved is the number everyone measures.

Instead of reconciling three versions of the same number at review time, your team agrees on it once, before the agent goes live, and every result after that is measured on the same ground.
Abandoned AI Agents Are Actually a Data Definition Problem
SnapLogic's respondents are not short on confidence. 85% say they are extremely or very confident in their ability to scale AI adoption.
But confidence in the agent does not settle what the agent changed. That answer lives in the data underneath it: which records, which definition, which owner.
Organizations that scale AI agents successfully do not just give agents access to more data. They make sure everyone measures the result on the same data.
Forward Deployed Watch: The Hottest Job in Tech Is Getting AI Into Production
Forward Deployed Engineer is now the second-hottest job in tech. Pave's latest hiring index, which tracks how fast each job family's share of new hires and headcount is changing, scored the role 97.0 out of 100 for Q3, just behind AI Engineer at 97.4. One quarter earlier it was only on Pave's emerging jobs watchlist.
An FDE is an engineer who embeds with a customer to get a product working inside its systems. The demand says something simple: buying AI is easy, getting it to work on your own data is not. SnapLogic's numbers above show what happens when nobody owns that last mile.
That is the gap Derisk360 was built for. It embeds pods of Forward Deployed Engineers, who build governed agentic systems inside your own cloud, alongside Forward Deployed Eval Engineers (FDEEs), who own trust after go-live: eval harnesses, red teaming, guardrails and drift monitoring, with audit-ready evidence. Every engineer is trained in the Derisk360 FDE and FDEE Factory, and Derisk360 reports go-live in under 12 weeks.
Warm regards,
Shen and Team
