Tax Teams Want AI Agents. Only 16% Can Use Their Own Data.
The agentic finance gap is a data ownership problem
Most organizations attribute slow AI progress in finance to the technology. The models are not ready, the vendors overpromise, the pilots take too long.
The larger problem is the data the agents are meant to work on. A new EY survey of 1,600 CFOs and senior finance and tax leaders found that 77% now expect AI agents to benefit the tax function, up from 21% a year ago. Yet only 16% consider their tax function highly effective at accessing, organizing, using and reusing its own data, and 64% name insufficient AI-ready data as the biggest barrier to advancing AI.
That gap lands on your finance and data teams as recurring work. Every filing, provision and reconciliation still depends on someone pulling numbers from the ERP, matching entities across systems and explaining why two reports disagree.
And it gets worse as agents arrive. A person can spot that a legal entity code changed last quarter. An agent reads whatever it is given, and it reads it at scale.
As AI agents take on finance work, every agent needs to know which data it can rely on, who stands behind it and which other agents it can trust. 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 Do Finance Teams Want Agents Their Data Can't Support?
Because the ambition and the data live on different timelines. Agents can be bought this quarter. Clean, connected finance data takes years of decisions nobody wrote down.
EY found that among leaders who say their function is not effective with data, 61% point to data silos and 52% to poor data quality. Those are not new problems. They are old problems that agents make visible.
Picture a tax agent asked to prepare a quarterly indirect tax return. It pulls invoices from one system, entity details from another and supplier records from a third. Two of them list the same subsidiary under different codes.
The agent produces a return.
Nobody can say which version of the entity it used.

The Cost Isn't Slower AI. It's More Manual Checking.
When the data underneath an agent cannot be trusted, the work does not disappear. It moves. Someone now has to verify every output the agent produces, usually the same senior people the agent was supposed to free up.
Before a finance or tax agent can be trusted, someone usually has to:
Confirm which system holds the authoritative record for each legal entity
Trace where a figure in a provision or return actually came from
Resolve duplicate supplier and customer records across ERPs
Find out who owns a field when its definition changes
Explain to auditors why last quarter's number moved
None of these tasks are particularly difficult. They are slow, they repeat every close, and the knowledge sits with a handful of people who are already stretched.
EY's own framing is telling. Marna Ricker, EY Global Vice Chair of Tax, said the challenge is transforming data, governance and talent capabilities, and that "ambition alone will not be enough."

How to Make Finance Data Agent-Ready in 30 Days?
You do not need to replace your ERP or rebuild your data warehouse to give agents something they can rely on.
Within the first month, focus on the entities every filing touches: legal entities, suppliers, customers and the chart of accounts.
Start by discovering what metadata you already have, where each of those records comes from and which reports, returns and models depend on them. Then confirm a named owner for each and agree which version is authoritative.
Next, expose this context through a single knowledge layer your analysts and agents both read from, so the answer to "where did this number come from" is never a meeting.
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 your agents work from the same trusted record your auditors expect.

Instead of spending months documenting finance data manually before every AI rollout, your team maintains the context once, and every agent, report and return inherits it.
Agent Readiness Is Actually a Data Ownership Problem
The finance leaders in EY's survey are not wrong to want agents. The work is repetitive, deadlines are tightening and the talent is scarce.
But an agent can only be as reliable as the least-owned record it reads. Most finance data has never had a clear owner, because a person could always fill the gap.
Organizations that scale AI agents in finance successfully do not start with the agent. They start by deciding who owns the data the agent will trust.
Warm regards,
Shen and Team
