AI agents for sales research

By Anna Fontanes | March 2026 | 10 min read

The single biggest bottleneck in modern prospecting isn't finding accounts. It's researching them.

You find a list of 100 accounts that match your ICP. That's the easy part. Database tools handle that now.

Then you sit down and spend 45 minutes per account finding their website, reading what they do, checking their job postings, pulling their Companies House filing, understanding their trajectory, identifying where they might have pain. You do this 100 times. That's 75 hours.

Most teams don't have 75 hours. So they skip it. They reach out to accounts they know nothing about, hoping something sticks.

That's where AI agents change the game. An AI agent can do the research in 2 - 3 minutes per account. Not the research that tells you they're a software company (the agent's too smart for that - they already know). The research that tells you why you should reach out right now.

What an AI agent research workflow looks like

Here's a concrete example:

Input:

Agent workflow:

The agent opens the company website. It reads the homepage. It understands: they're a B2B SaaS company selling marketing automation. Their target customers are mid-market companies. They're positioned on ease of use and customer support.

The agent checks their LinkedIn job postings (if available). It finds: they've posted three jobs in revenue roles in the last quarter. Two SDRs, one sales operations person. That suggests they're scaling revenue.

The agent pulls their Companies House filing. It finds: they filed accounts six months ago showing £12M revenue (up from £8.5M the prior year). That's 41% growth. They also just appointed a new Finance Director four months ago.

The agent connects the dots. Revenue growing at 40%+ annually. Scaling the sales team. Bringing in new financial leadership. These are signals of growth mode and expanded budget. They're probably evaluating sales efficiency tools.

Output: Structured research brief

"Example Ltd is a B2B SaaS marketing automation platform. They've grown revenue 41% year-on-year to £12M (filed accounts from six months ago). They appointed a Finance Director four months ago, suggesting preparation for growth or capital raise. They're actively scaling the sales team (posted 3 roles in revenue functions in last quarter). All signals point to a company in growth mode evaluating sales infrastructure and efficiency tools. Outreach angle: help them convert growing headcount into proportional pipeline growth."

You've got a one-page brief. You know exactly why you're reaching out and what angle to use.

Why this matters more than faster email copy

When sales tools first integrated AI, the assumption was: AI makes us faster. Use AI to write emails faster. Personalise at scale.

That's helpful, but it's not where the real value sits.

The real value is: AI accelerates the work that actually drives results. Not copy, but context. Not volume, but relevance.

An agent doing research doesn't just save you time. It changes what you can do with that time.

Without AI: You have 10 hours per week to research accounts. You can research 10 - 15 accounts at depth. You reach out to 15 people per week.

With AI: You can research 100 accounts. You can reach out to 100 people. But more importantly, you're reaching out with genuine context for all 100 instead of researching 15 thoroughly and blindly emailing 85 others.

That's not slightly better. That's a different category of prospecting.

The building blocks: what the agent actually does

An effective research agent needs access to three sources of data:

1. Website and public company information

The agent reads the company website. It understands their positioning, who they claim their customers are, what problems they solve, what their growth narrative is. This isn't just extracting company description - it's interpreting it. It can tell the difference between "we serve mid-market SaaS" and "we serve mid-market SaaS that use HubSpot," which are very different.

2. Job posting data

Job postings tell you what a company is prioritising right now. A CFO hire signals financial rigor or prep for capital. Multiple sales hires signal revenue scaling. A Head of RevOps signal tool standardisation and process maturity.

The agent doesn't just see "they hired someone." It understands what type of hire suggests what type of growth. A Director of Data Analytics hire is different from an SDR hire, which is different from a VP of Product hire. They all signal different things about company trajectory and priorities.

3. Companies House financial data and director appointments

This is the big differentiator for UK prospecting. Companies House filings give you revenue, cash position, debt, growth rate, and director appointments. You can see when someone was appointed (exact date in some cases). You can see revenue growth over multiple years.

An agent that can read Companies House data can answer questions no database can:

Why this workflow works for high-touch prospecting

The research brief becomes your outreach foundation. You don't write blind. You write based on concrete signals.

Instead of "Hi [Name], I noticed you work at [Company], we work with similar companies" - which is obviously templated - you write:

"Hi [Name], I noticed Example appointed you as Finance Director four months ago. The company's grown 41% year-on-year to £12M (based on latest filed accounts), and you're now scaling the sales team - you've posted for 3 revenue roles in the last quarter. That's classic growth mode, and typically companies scaling like that are evaluating tools to help convert headcount into proportional pipeline growth. We work specifically with companies at that stage. Worth a conversation?"

That's different. That's someone who's done 15 minutes of research and clearly knows something about the company's trajectory.

The response rate difference is material. A templated message might get a 1 - 2% reply rate. A contextual message based on real signals gets 15 - 30% reply rates from relevant accounts.

Multiply that by 100 accounts and you're looking at 15 - 30 conversations instead of 1 - 2.

The workflow in practice

Monday morning: You give the agent a list of 50 companies and your 30-second description of what you sell. "We're a sales intelligence platform helping B2B SaaS teams find and prioritise accounts without needing a RevOps team."

By Tuesday morning: You have 50 one-page research briefs. Each one tells you what the company does, what signals they're showing, and why you should reach out.

Wednesday - Friday: You reach out to all 50. You're not reaching out blind. You're reaching out with real context.

This is completely different from the spray-and-pray workflow:

Monday: Paste 5,000 names into an email tool. Click send.

The first workflow involves 4 hours of work and potentially converts 8 - 15 of your 50 targets into conversations. The second converts maybe 5 - 10 of your 5,000 into conversations.

One team is working smart. One is working hard.

What happens after the brief

The research brief is your foundation. You still write the email. You still make the call. You still need to be sharp in the conversation.

But you're not starting from zero. You're starting from "I know this company, I know why they might need us, and I have a specific angle."

That changes everything about the conversation. You ask better questions. You listen for specific things. You don't get lost in generic "tell me about your current solution" territory.

Building this into your process

If you're a team of 5 - 20 sellers without RevOps support, this workflow is game-changing. You can't hire someone to do research. But you can use an agent to do it.

The agent doesn't replace your sales expertise. It augments it. It handles the grunt work - reading websites, scanning filings, cross-referencing job postings. You handle the human work - writing contextual emails, having conversations, understanding nuance.

That's a much better division of labour than "you research 10 accounts yourself and email the other 90 blind."

AI agents work better when they have better data

Firmbase integrates with research agents (via our MCP integration with Claude) so agents can pull complete account data, Companies House filings, growth signals, and director information in real time.

Start your free trial or get in touch to set up MCP integration for your AI research workflow.

FAQ

Q: Do I need to manually feed the agent company names, or can it pull from a list automatically?

A: That depends on the platform. Most modern agents can pull from a database or spreadsheet automatically. You give them a list, they run the research workflow on all of them.

Q: What if a company doesn't have Companies House data (e.g., it's a subsidiary or private company)?

A: The agent works with whatever data exists. It reads the website, checks job postings, and understands what it can from public sources. For subsidiaries, it might find parent company information instead. It's not perfect, but it's still more informative than no research.

Q: How accurate is the agent's interpretation of company data?

A: The agent is as accurate as the data sources it reads. If Companies House shows revenue, that's official. If the website says they serve mid-market customers, that's accurate (assuming they're not lying on their website). The agent synthesises this data accurately. It doesn't hallucinate or guess.

Q: Can the agent identify whether a company is actually in my ICP?

A: Yes. You give it your ICP criteria ("B2B SaaS companies with revenue between £2M and £10M in revenue"), and it can filter based on that. It reads the company information and tells you if they fit.

Q: How do I actually use the research brief in my outreach?

A: You read the brief and write your own email. Don't copy-paste the brief into the email (that's weird). Use it as your research foundation. Reference the specific signals in your own words.

Q: Isn't this just another way to do the same research I'm already doing?

A: No. You're not doing this research now. You're either not doing it (spray and pray) or doing it manually (which doesn't scale). The agent makes precision prospecting actually doable at the scale most teams need.

Author Bio

Anna Fontanes is a revenue operations consultant who has built account scoring and ICP frameworks for UK B2B sales teams across SaaS and professional services. She specialises in making structured prospecting work for teams without dedicated ops resource.