IDC: AI Agents Will Run Most Processes by 2029. Which Data Matters?
More data won't help agents that can't tell what's critical
Most organizations attribute disappointing returns on data to a shortage of tools, talent or AI. Buy a better platform, hire more engineers, add agents, and the value will follow.
The larger problem is that few can say which of their data actually matters. On October 27, IDC predicted that by 2029, 60% of the world's 2,000 largest enterprises will run most of their business processes AI-first, up from 13% in 2026. In the same FutureScape, IDC predicted that by 2028, 60% of enterprises will fail to convert data into measurable business value, because data volumes outpace their ability to define and govern that value.
That leaves data teams with a recurring job nobody planned for: deciding, process by process, which tables an agent can rely on, which it should never touch, and who answers for the difference. Today that judgment usually lives in the heads of a few senior engineers.
And it gets worse as agents spread. IDC's survey data shows enterprises have already put agents into an average of 11 production workflows. Every new workflow adds datasets someone has to vouch for, while the data itself keeps growing faster than anyone can rank it.
As processes go AI-first, agents will need to find, verify and work with agents outside their own company, not just the data inside it. That is what we are building at OpenMap: The Agent Network, the connective tissue for the world's AI agents, starting with every person and every local business. We are raising a £5M seed. Meet us at openmap.ai/investors.
Why Does More Data Produce Less Value?
Because volume is not the same as value. Most data estates grow by accumulation: every project adds tables, every migration leaves copies, every team builds its own extract.
Picture a finance team handing month-end close to an AI agent. The warehouse holds tens of thousands of tables. Perhaps a few hundred actually feed revenue, cost and cash. Nobody has written down which few hundred.
The agent can read all of them.
It cannot tell which ones the CFO would sign off on.
This is not a data volume problem. It is a data prioritization problem.
Everyone Thinks They're Ready. Their Data Disagrees.
Leaders tend to rate their own readiness highly. In Precisely's 2026 State of Data Integrity and AI Readiness research with more than 500 data and analytics leaders, 88% said they have the data readiness AI requires. Yet 43% named data readiness as a major obstacle to their AI goals.
Both can be true at once. The data is ready for the reports people already know how to read. It is not ready for an agent that has to act on it without asking anyone.
Before a process can safely go AI-first, someone has to:
Identify the datasets that process actually depends on
Rank them by the decisions and money that flow through them
Name an owner for each critical dataset
Agree on one trusted version of each key entity: customer, product, supplier
Map what breaks downstream when one of those datasets changes
None of these tasks are particularly difficult. The problem is that they are done once, for a single project, and go stale as the next ten processes are automated.
How to Find the Data That Matters in 30 Days?
You do not need to inventory every table in your warehouse to do this.
Within the first month, focus on the two or three processes most likely to go AI-first next: the ones your business leaders are already asking agents to run.
Start by tracing each process back to the datasets that feed it, and resolve duplicate and conflicting records for the entities it touches. Then attach an owner and a policy to every dataset on that short list.
Next, expose this context through the same channels your agents and engineers use to find data, so "critical" is a label they can see, not something they have to ask around about.
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 data behind each AI-first process is known, owned and current.
Instead of spending months ranking assets in spreadsheets, your teams can see which data each AI-first process depends on and put their effort there first.
Low Data Value Is Actually a Prioritization Problem
IDC's two predictions point at the same gap. Enterprises are about to hand more decisions to agents just as their ability to say which data deserves trust falls behind.
More data will not close that gap. Neither will more agents.
Organizations that scale AI-first processes successfully do not govern every dataset equally. They know which data matters, who owns it and what depends on it.
Warm regards,
Shen and Team