Event Recap: Modern Consensus Management at CIRI Senior IRO Leadership Day

Ask any IRO about consensus and you'll hear the same frustration. A group of analysts builds forecasts on your business, each with their own assumptions. When results land below those forecasts, the Street says you missed.

That tension opened our session, Modern Consensus Management: Shifting from Grunt Work to High-Stakes Judgment, at the CIRI Senior IRO Leadership Day on October 7, 2026. Our founder and Chief Strategy Officer, Darrell Heaps, lived it firsthand as a public company CEO. He walked senior IROs through our research on working with the sell side, and showed how agentic AI can take on the grunt work so IR teams can spend their time on judgment.

Here's what he covered.

Consensus Is a Credibility Operation

There's an old line in IR: it's a three-month job that repeats four times a year. To understand that cycle, Q4 spent 25+ hours in interviews with senior IR, FP&A and CFO professionals, then tested prototypes back with the same group. One theme came up again and again. Every interaction with the sell side builds credibility or spends it.

That credibility belongs to the IRO as much as to management. As one retired SVP told us, "Every single thing we do in investor relations is to maintain or improve your credibility." Once analysts start doubting the IRO, they go looking for information elsewhere. Trust takes years to build and seconds to lose.

Academic research backs this up. Companies with tight consensus that consistently meet expectations tend to see stronger returns than those that miss.

Darrell stressed that dispersion, or how tightly analyst estimates cluster, carries as much weight as the headline number. Each estimate is an analyst pitching a view to their clients. A wide range means the Street is telling several different stories about your company.

Leaning on Aggregators Hands Your Story to Someone Else

So how do you shape a model you can't comment on directly? You can't call an analyst and say their revenue number is 10% too high. Disclosure rules put that off limits. And with 20, 30 or even 60 covering analysts, one outlier can drag consensus the wrong way.

Many teams fall back on third-party aggregators for their consensus view. Most people in the room had seen an aggregator number and wondered where it came from. Darrell's advice was to own consensus yourself. The best companies collect every model and build their own aggregate. That tells them which analysts move the number.

Doing it by hand brings its own problems. The deck framed them as three hidden risks:

• Siloed execution. Models get consolidated in spreadsheets, often by several people. Earnings week stacks heavy workloads on short timelines, which is when errors creep in.
• Internal modeling mismatch. IR and finance can end up on different baselines. That makes guidance hard to test against live expectations, and circulating draft forecasts create disclosure exposure.
• The aggregator blind spot. Third-party feeds can carry stale estimates or unadjusted one-offs. When those hit a headline, IR is left defending the narrative after the fact.

Consolidation tools help with the heavy lifting. Yet a merged spreadsheet still hands you the haystack and leaves you to find the needle. As one slide put it, "When consensus data lives everywhere, confidence lives nowhere."

Each Phase of the 90-Day Cycle Needs Its Own Playbook

Working with the sell side runs the full 13 weeks. Meeting analysts after the call covers one slice of it. Our research split the cycle into three phases. Each has grunt work AI can take on and judgment calls that stay with you.

Pre-quiet period: anchor consensus before lockdown

This is your open window. You have time to dig into models, find the outliers and learn what's driving them. Done by hand, it takes hours of aggregation before you even know which analyst to call.

Darrell showed a prototype agent that does this work in the background. It plots every analyst's estimate by line item and flags the biggest outliers in red. Click into one and you see a waterfall from internal consensus to that analyst's estimate, along with their full model, thesis and likely questions.

A compliance agent sits alongside it. It checks every suggested talking point against what you've already disclosed in press releases and transcripts. Anything new gets flagged as a Reg FD risk before it reaches your meeting notes.

• AI takes: line-item driver analysis across every model, flags for analysts above or below guidance, a dispersion summary and 30-day revision momentum.
• You keep: which outliers to engage and how, tightening messaging within disclosure rules, and deciding when a gap is noise and when it's a signal.

Quiet period: prepare with more than transcripts

Once the quiet period starts, the work turns inward. You're weighing what you promised against what you'll deliver, with the Street's expectations in between.

Most of us have seen AI predict analyst questions from past transcripts. The difference shows up when the data set adds research and models. Then an agent can predict specific technical questions tied to each analyst's estimates.

In the demo, run on historical data from a former public company, the agent ranked question themes by frequency and sentiment. It showed which analysts probe hardest and closed with tiered questions ranked by probability, each with a suggested approach.

• AI takes: a promised vs. expected vs. delivered review, a peer transcript and results scan, predicted analyst questions and a tone check against prior calls.
• You keep: framing the guidance narrative, aligning the CEO and CFO message, preparing for the hardest questions and setting callback priority.

Post-earnings: correct models before they lock

Then comes the gauntlet. At Darrell's first earnings call as a public CEO, he had eight analyst calls lined up right after it. Each 15-minute meeting is your chance to point to what you disclosed and where a model drifts from it. Miss that window and the note goes out with the error baked in.

Results and guidance can shift right up to disclosure, and every change ripples through your prep. An agent can rebuild that prep the moment the call ends. The prototype showed a heat-mapped grid comparing each analyst's estimate with consensus and guidance, line by line. Click an analyst for their likely questions and a meeting strategy, including topics to steer clear of.

• AI takes: reported results vs. each analyst's prior model, separating mechanical errors from assumption errors, root cause by line item and reconciling third-party data feeds.
• You keep: the objective of each callback, who calls whom and in what order, owning the error as the company's, and telegraphing metric changes one to two quarters ahead.

Research participants were clear on one point. They want the topics to focus on, and they'll choose the words themselves. Darrell compared it to GPS. It gives you directions, but if it says turn right into a lake, you don't drive in.

Generic AI Works for Documents. Consensus Needs Governed Data.

Darrell drew a clear line between chat-based AI and agentic workflows. The first wave gave you a chat box and asked you to become a prompt engineer. Agents run in the background, often overnight, and hand you output you can act on. You can still ask follow-up questions, but the starting point arrives ready.

For high-stakes financial data, the foundation under the AI decides whether you can trust it.

Clean models and governed data are what make AI output safe to bring into an analyst meeting.

AI Scales the Execution. The IRO Leads the Judgment.

The session landed on a new division of labor. Infrastructure handles continuous monitoring and outlier detection, with expert analyst validation behind every number. The IRO protects management's credibility and frames the story the Street hears.

Q4’s consensus management puts that division of labor into practice. It brings every covering analyst’s model into one view and keeps the Street’s numbers lined up with your internal forecasts. Q, our AI, works across that data so you know which analyst to call and what you can safely say. More agents are on the way, built alongside IR teams who run this cycle every quarter.

Want the full story, including Darrell's walkthrough of each demo? Listen to the session recording.