How I Built The Company Research AI Agent I Always Wanted

I was ready to launch a new agent at 1 am last week.

Everything was working. The agent was doing its thing. But I decided to wait until morning -- mostly because I've shipped enough products to know that launch decisions made at 1 am are hit or miss.

So I launched the CompanyResearch.ai agent the next day instead.

I built this for myself because I'm constantly looking up information about companies -- startups I'm curious about, potential investments, HubSpot customers/partners, competitors in different spaces. Companies people have worked for previously. It's one of those tasks I do several times a day, every day. I got tired of jumping between ten different tabs to piece together what I was looking for.

Now, some of you might be asking yourself: “Wait, couldn’t you just do this with ChatGPT deep research”? Yes…and no. You can do some of it with a really well crafted prompt or CustomGPT. But a lot of the information I’m interested in is not available on the public web in a convenient way. And when I do try to pull this information using an LLM, unless I ground it with actual data, the hallucination rate is too high.

Anyways, trust me, if I could have gotten away with just writing a well-crafted prompt, I would have done it. I’m lazy that way.

In any case, the response was wild. Within 48 hours, I've shipped two major updates based on user feedback (one at 2 am, another at 3 am -- I clearly have a pattern with late-night shipping).

So today, I want to break down:

• What the Company Research agent does

• Who this is particularly useful for and quick tips

• Why I'm iterating so quickly (and what’s next)

— @dharmesh

What the Company Research agent does

The Company Research agent pulls from multiple sources -- including several proprietary/paid databases -- to create a comprehensive report on the company you're looking to explore.

You can research with this agent in two ways:

1. Direct lookup: Drop in the targeted company domain or name (hubspot.com, or openai.com or whatever company you're curious about)

2. Category search: Instead of searching for a specific company, you can also search by category and criteria. For example, "HR SaaS companies," or "Law firms in Boston" to get a list and then choose a company to research from there.

3. While it works on pulling the data (and there's quite a bit happening under the hood), you get quirky loading messages like "Working hard so you don't have to."

After about 30 seconds, here's what you get:

• Company Overview: Covering things like its website, location, founding year, industry, employee count, and annual revenue (for public companies, includes revenue multiples and revenue per employee)

• Product Summary: What the company does and what products they offer

• Homepage Analysis: Key messaging and positioning from their site

• Competitors: Who they're competing against

• Technology Stack: One of my favorite sections. Shows the complete tech stack -- analytics tools, advertising platforms, JavaScript frameworks, CDN services, security tools, CMS platforms, and more.

Plus: there are additional sections, including search keyword analysis, hiring and job openings, HubSpot-specific info (if they're a customer), web traffic metrics, key people (with inline search to find specific executives), and recent news and press.

If this is too much, you can also use the Customize tab to control exactly what appears in your reports. You can drag sections between "Included" and "Excluded" lists, reorder them, or even add custom questions specific to what you're curious about.

You can also save companies, push generated reports directly to HubSpot, or share links to them with your team (no login required).

The agent currently runs on GPT 5.1 (shipped with the second update at 2am), which I think made a noticeable difference in both speed and accuracy compared to the initial launch version.

Who is this agent for?

I started by building this agent for myself, but it turns out that many people have the same problem I do.

Sales people have been using it to research prospects before calls -- instead of spending 20 minutes jumping between LinkedIn, the company website, SimilarWeb, and G2 reviews, they get everything in one place in 30 seconds.

Founders and entrepreneurs have been using it to look-up competitors.

Partnership teams have been using it to evaluate potential partners and understand their tech stack, market position, and recent momentum.

Investors have been using it for preliminary company research before deeper diligence -- getting a quick sense of the business model, competitive landscape, and growth signals.

Customer success teams have been using it to understand their accounts better -- especially useful when a new account lands on your desk and you need to get up to speed quickly.

And, people looking for new job opportunities are using it to research a company they’re applying to. Nothing impresses an interviewer more than having really done your homework and learning about the company.

Honestly, it's for anyone who's tired of manually collating data from different sources just to answer basic questions about a company. That was me, and based on the response so far, that's a lot of other people, too.

Why I'm iterating this fast (and what’s next)

Here's the thing about building products: you can spend weeks planning the perfect launch, or you can ship something useful and iterate based on real feedback from the community.

I chose the latter.

• Day 1: Launched the core product

• Day 2: Bug fixes, UX improvements, better people search, high-level G2 stats

• Day 3: Upgraded to GPT 5.1, added Tech Stack section, improved accuracy

• Day 7: Share it with my entire newsletter community (that’s you!)

The first version was far from perfect -- it had bugs, the UX could be better, and some sections didn't work quite right.

But it was useful. And that's what matters.

Instead of spending another month perfecting features that users might not even care about, I'm letting real usage guide what to build next.

In fact, some of the best ideas came directly from the first 48 hours of feedback -- like showing the complete tech stack and adding inline search for key people.

As I continue building, a few things are on my radar for the next update:

1. A SmartNote feature: Automatically generates a customized summary and adds it to HubSpot on the company record, with a link back to the full report. For sales teams who want the research logged directly in their CRM.

2. More customization: Deeper control over what sections appear and how they're formatted.

3. Quarterly earnings analysis: For public companies, pulling insights from earnings calls to understand recent performance and strategy shifts.

The roadmap will keep evolving based on what users need… so I’d love it if you let me know what you’d like me to add next! I read every poll response and reply to these emails.

Try it yourself

If you're doing any kind of company research -- whether you're in sales, partnerships, investing, or just curious about the competitive landscape -- you can try this agent via agent.ai or at companyresearch.ai.

It isn't perfect (nothing is after a week), but it's definitely useful. And it's getting better every day.

And don't forget: Keep the feedback coming. I'm listening, and I'm shipping updates as fast as I can build them.

—Dharmesh (@dharmesh)