A year ago, 6% of IR teams were actively using AI. Today that number is 42%.
That’s not a trend. That’s a shift. And it happened fast enough that a lot of IR professionals are navigating it without a map – unsure which tools are worth their time, which claims to believe, and where the real risk lies.
What AI actually is
Most AI tools you’ll encounter today are large language models – systems trained on enormous volumes of text to recognize patterns and generate useful responses. Think of them as extremely sophisticated pattern-matchers. They’re very good at tasks that involve synthesizing, summarizing, or drafting based on existing information.
What they’re not: they don’t reason, they don’t verify facts independently, and they don’t understand the strategic context behind your company’s story. They process what they’re given and generate what’s statistically most likely to be helpful. That distinction matters a lot for how you use them.
Where AI earns its place in IR
If you’re getting real value from AI right now, it’s probably in a specific category of work: information-dense, time-intensive tasks where the bottleneck is processing volume, not making judgement calls.
Transcript analysis leads the list – 35% of AI-using IR teams rely on it for earnings call and transcript review, making it the most widely adopted use case in the profession. Scanning a 60-page peer earnings transcript to find out how management addressed margin pressure, what guidance language they used, which analyst questions drew the longest pauses – that's work AI can compress from 45 minutes to five.
Communication drafting (27%) and meeting preparation and research (20%) round out the top three. All three share the same pattern: they reduce information-processing time while leaving the strategic decisions – what to say, how to say it, how to read the room – to you.
Where AI adoption drops sharply is in judgement-heavy work. Only 5% of teams use AI for investor targeting, and just 11% for sentiment analysis. Those are not easy tasks to automate, and the low adoption reflects that IR professionals know the difference between a time-saving tool and a strategy replacement.
The AI features already in your Irwin + FactSet Workflow
Most of the conversation about AI in IR focuses on what’s coming – the next model, the next integration, the capability that’s six months away. Less discussed: what’s already available and what it actually does day-to-day.
If you’re using Irwin and FactSet today, you have access to a set of AI functionality designed specifically for IR and Corporates workflows. Here’s what’s in the platform now:
Irwin AI Summaries – Built from your own CRM data
Every interaction you log in Irwin – meeting notes, call summaries, investor feedback – is analyzed automatically by Irwin’s AI Summaries feature. It surfaces key themes, flags sentiment shifts, and builds context for your next engagement without you having to search through months of entries manually.
In practice, this means logging your notes after a meeting – what they asked, what concerned them, what got their attention – and having Irwin surface those themes the next time that investor comes up. Walk into a follow-up six months later knowing exactly what they raised last time. Your notes become institutional memory rather than a buried CRM entry.
FactSet Transcript Assistant: Ask your transcripts a question
Transcript analysis is one of the most time-intensive tasks in IR prep. Scanning competitor earnings calls, peer disclosures, or analyst day transcripts line by line is slow – and it’s exactly the kind of work where detail gets missed under time pressure.
FactSet’s Transcript Assistant is a conversational AI tool that lets you query transcripts directly. Ask it what guidance commentary appeared in a peer’s Q3 call. Ask it how management talked about capex over the last three quarters. You get a sourced, specific answer in seconds – not after 45 minutes of reading.
And because it’s built on FactSet’s data – with citations – you know exactly where the answer came from. Every response traces back to the underlying document.
AI-Powered Document Search: Find what you need, faster
Beyond the Transcript Assistant, FactSet's AI-enabled Document Search works across the full universe of unstructured content — earnings calls, filings, news, research, and more. Instead of reading to find the relevant section, you ask a specific question and surface what matters, with every result source-linked and traceable back to the original document.
The feature launched in broad beta to more than 85,000 users in early 2026 and is rolling out globally through spring.
There are a few things worth knowing about how it actually works:
The natural language agent — powered by FactSet Mercury — lets you ask questions directly and get synthesized answers drawn from across FactSet's content. Ask it to summarize a company's product strategy broken down by segment. Ask it how a specific company has talked about guidance over the past four quarters. You get a structured answer with citations, not a list of documents to read yourself.
The comparison grid lets you benchmark companies and timeframes side by side — useful for peer analysis during earnings prep, or any time you need to understand how a narrative has evolved across your coverage.
The AI-curated homepage surfaces priority topics automatically, so you're not starting from a blank search every time you need to get up to speed on a company or sector.
The common thread across all of it: results are drawn exclusively from trusted sources, every answer is citable, and nothing leaves you guessing where it came from. In an IR context — where what you say to investors needs to be defensible — that auditability isn't a nice-to-have.
The common thread: auditability
What distinguishes the AI built into Irwin and FactSet from general-purpose AI tools isn't just that it handles IR-relevant data — it's that every answer is traceable. When you surface an insight and present it internally, you can show exactly where it came from.
That's not a minor point for IR teams. It's the difference between an AI answer you can stand behind and one you have to qualify.
