AI in Investor Relations: Disclosure, Q&A and Consistency

Investor relations is a consistency discipline operating under legal constraint. The same facts must appear identically across a release, a presentation, a filing, and a call, and any statement about the future carries weight. Much of the work is checking and assembling, which is automatable; the judgement about what to say is not. This guide covers the split, drawing on FISTA Solutions' AI agents work with executive functions. It complements how to build a board reporting assistant and ai guide for heads of investor relations. This article is general guidance, not legal advice.

What does consistency checking catch?

Differences nobody intended. A figure in the presentation that does not match the release. A characterisation of a segment's performance that differs from the filing. Language about a risk that has changed from last quarter without anyone deciding to change it.

These arise because materials are assembled by several people under deadline, and they are noticed — by analysts, by the financial press, and occasionally by regulators. Automated comparison across the material set catches them before publication.

Activity	Automatable	Human required
Cross-material consistency checks	Yes	Resolution decisions
Prior-period language comparison	Yes	Judgement on changes
Question anticipation	Yes	Answer preparation
Transcript and coverage analysis	Yes	Interpretation
Forward-looking statements	No	Management
Responses to investor enquiries	Constrained	IR and legal

How does question anticipation work?

By assembling from what already exists. Prior call transcripts show what analysts ask this company. Peer calls show what they are asking the sector. Analyst notes show current concerns. The results themselves indicate where questions will land.

Producing a likely question list with the company's prior answers and the current position attached lets preparation focus on the genuinely difficult questions rather than on constructing the list, which is where most of the preparation time currently goes.

What does transcript analysis add?

Scale. Reading every peer call, every analyst note, and every prior transcript to track how focus shifts is beyond what an IR team of two or three can do manually.

Knowing that a topic has appeared in four consecutive peer calls, or that analyst attention has moved from growth to margin, informs messaging and expectation management before it becomes a surprise on a call.

Why must forward-looking statements stay human?

Because they carry legal consequences and represent management's considered view of the future. Generated guidance language is a statement of intent that nobody deliberated over, in a context where statements are relied upon and litigated.

The boundary should be explicit in the system: it may assemble prior language and flag inconsistency, and it may not produce a statement about expected performance.

What constrains answering investor questions?

Selective disclosure rules. Conveying material non-public information to some investors and not others is a regulatory problem regardless of the channel or the mechanism.

Any automated response capability must therefore be confined to already published material, with a hard boundary and a clear escalation for anything else. That constraint is severe enough that many IR teams reasonably decide not to automate outward responses at all. See what is a guardrail policy.

What about shareholder and analyst research?

Assembling who holds the stock, how positions have changed, and what coverage says is useful preparation work. Where data is licensed, the licensing terms govern what may be done with it, which is worth checking before building a pipeline on top of it.

Who should own it?

The IR function, with legal holding a veto on anything touching disclosure. IR systems fail at precisely the seam between usefulness and regulated statement, and that seam needs an owner who understands both sides.

How is it evaluated?

Questions anticipated that were actually asked, inconsistencies caught before publication, preparation time per results cycle, and corrections issued. Materials produced measures effort rather than quality.

What goes wrong?

Generated forward-looking language. Automated responses to investor questions. Consistency checking limited to numbers rather than characterisations. And treating a licensed data feed as freely usable in a derived product.

What does it cost to run?

Low, since volume is a handful of cycles a year with concentrated preparation. The value is in preparation quality and error prevention rather than in cost saving, which is worth stating clearly to a sponsor expecting an efficiency case.

What should you do first?

Take your last results pack and check every number and characterisation across the release, presentation, and filing. Whatever you find is what an automated check would have caught, and in most cycles there is something.

How does this apply outside results cycles?

To the steady work between them: maintaining a current fact base that every draft is checked against, tracking what has already been said publicly on a topic, and keeping the question list current as coverage evolves. That continuity is what makes each cycle's preparation shorter, and it is the part most often abandoned once a results period ends.

How FISTA Solutions helps

FISTA Solutions builds investor relations systems with cross-material consistency checking covering characterisations as well as figures, question anticipation assembled from transcripts and coverage, and sentiment tracking across peers, while forward-looking statements and outward responses stay under IR and legal control, through AI agents, AI enablement, and forward deployed engineers. The record behind the approach is 150+ projects for 50+ companies with 47% efficiency gains.

To prepare better and publish consistently, message FISTA on WhatsApp, or read how to build a board reporting assistant.