The Digital Developer Blog - Blogs by Computan

Why Your CRM Never Did What You Bought It For, Until AI Connected It

Written by Simranjeet Singh | September 4, 2026 at 11:00 AM

 

Insights from an Unwatchable episode with Joe Jerome (CRO, Computan) and Sajeel Qureshi (CEO, Computan)

TL;DR

  • Most CRMs sit half used because logging calls, emails, and follow ups by hand is a habit almost no sales or marketing team keeps up consistently.
  • Connecting AI tools like Claude to a CRM like HubSpot closes that gap because the AI can document activity, draft follow ups, and flag who has gone quiet without a person doing the manual entry.
  • Joe Jerome put it directly: "This is the golden opportunity is to marry the AI to the CRM. So the CRM is finally now complete. It's finally doing what people bought it ten years ago to do."
  • Sajeel Qureshi framed the CRM's core function this way: "That's what a CRM is, right? I mean, it'll tell you anything you want that you've already told it."
  • Computan's stack for this workflow is HubSpot, Claude, Notion, and N8N, plus Fireflies for call transcription.
  • Agencies that try to build this kind of AI and CRM integration alone often underestimate the cost and complexity, according to Sajeel Qureshi: "there's just so much to it that goes into this doing it well, doing it properly."
  • Computan has completed more than 350 integrations and says the real advantage going forward will belong to whoever runs the most comprehensive AI and CRM setup, not the flashiest one.

The Problem: CRMs Only Know What You Tell Them

A CRM is only as useful as the data inside it. That sounds obvious, but it is the reason so many HubSpot and Salesforce portals turn into digital junk drawers. Sales reps forget to log calls. Marketing teams forget to update deal stages. Nobody goes back in after a meeting to write down what was decided.

Sajeel Qureshi described the mechanic behind this problem clearly: "CRM will tell you if you tell a CRM that I talked to Joe at two o'clock on Wednesday and we talked about this and then I'm gonna talk to him on Thursday about the same thing again later on. If you go back to CRM, it'll tell you, hey, yeah, you talked to Joe on Wednesday and you're supposed to talk to him on Thursday, it'll tell you all that stuff once you tell it what it needs to tell you."

The catch is the "once you tell it" part. That is where most CRM adoption breaks down.

What Changes When AI Gets Plugged Into the CRM

Joe Jerome walked through a real example of using AI to solve a recurring sales problem: people who stop responding to email. He explained his process like this: "I basically gave it some examples of what I write, whatever. I said, look, go through all my emails and tell me everyone that I either failed to get back to on my own or people who are failing to get back to me, right? Who ghosted me, right?"

The AI scanned sent and received emails, cross referenced who still needed a response, and drafted starter follow ups based on his usual tone and past sales training. Joe described the result: "I had about fifteen people in drafts. I had to slightly, I'm not gonna have AI do all of it, that's always a bad idea, and I had to modify most of them, but at least it was that start and that got it going. And I was able to just fire these off like within fifteen minutes."

That single workflow illustrates the bigger shift both hosts kept circling back to: AI does not replace the CRM, it feeds it and activates it.

Key benefits of connecting AI to your CRM

  • Follow up emails and ghosted leads get surfaced automatically instead of relying on memory
  • Call transcripts and meeting notes get logged without a rep manually typing a summary
  • Sales meetings get shorter because the AI already prepared the status of every deal
  • Bulk record management (reassigning contacts, updating fields) becomes possible at scale
  • Data that used to live only in someone's head becomes searchable, trackable, and reportable

Documentation Without the Manual Labor

Sajeel Qureshi pointed out that the labor of documentation is exactly what most teams skip. "Having context is important. So all this data should live somewhere. So you do have to, somebody has to document this somewhere. May as well be a robot documenting a lot of this stuff somewhere," he said. He added that logging calls back into a CRM "is a habit that most sales reps don't have. Marketing teams don't have it. Some do, obviously, not everybody, but most don't have this habit of keeping your CRM up to date all the time."

Joe Jerome said this shift changed his own CRM usage in a very concrete way: "With AI, I noticed that I'm using HubSpot even more now. Like I'm using the call feature a lot more, I'm using call transcripts a lot more, rather than calling from my cell phone, which I normally would do, I probably wouldn't log."

The practical takeaway is that AI does not eliminate the need for a system of record. It makes the system of record actually get filled in.

What a Connected Sales Meeting Actually Looks Like

Joe Jerome described how his team now runs sales meetings using AI prep that pulls from email, call transcripts, and internal chat channels. "It checks all of my emails, all of your emails that would be logged, all of my transcripts, all of your transcripts, all of the company transcripts. It checks our internal project channels and client channels in our chat. It goes through all these things. And it might also cross reference HubSpot as well," he explained.

That prep feeds a short, focused standup. "We have a fifteen minute sales meeting where we review that list and I go, okay, do you got this? I got this, who got, does Minaj need to send an invoice? Does Brat need to onboard someone? Whatever," Joe said.

The loop closes the next day, because the AI checks whether the assigned tasks actually happened. As Joe put it: "The AI then says, did Joe do this? Did Sajeel do this? Okay, well we did some of them, but not all."

The Tool Stack Behind It

When asked to name the core setup, Joe Jerome summarized it simply: "So when we talk about our main tool set, we're talking basically HubSpot, Claude, Notion, N8N." Sajeel Qureshi added Fireflies for note taking and standard messaging tools like Slack for communication.

Layer Tool Used Purpose
System of record HubSpot Deals, contacts, tasks, historical activity
AI layer Claude Drafting follow ups, prepping meetings, cross referencing data
Knowledge base Notion Team documentation and shared context
Automation and integrations N8N Connecting systems and custom workflows
Call intelligence Fireflies Transcribing calls for AI to reference later

Why Most Agencies Cannot Just Build This Themselves

Sajeel Qureshi was candid about the gap between what agencies think they can do internally and what actually happens. "People will think that they can do it on their own, they can do it on their time, on their budget, those sorts of things, and they can and they can't, right? I mean there's just so much to it that goes into this doing it well, doing it properly. You know, the data needs to be right, needs to be mapped properly, all that needs to be in sync and it's just not easy," he said.

Joe Jerome pointed to the custom integration piece specifically as the part most teams underestimate. "Most people are missing the custom integration piece because often people who write custom integrations charge far too much money and make it cost prohibitive," he explained, adding that Computan has "a model where we are able to deploy these and deliver these at a third of the price of the going competitive rate."

He also pointed to volume as a credibility marker: "We have the reps, you know, we've done over three hundred and fifty integrations."

Common gaps agencies run into when trying to DIY this

  • Underestimating how much time it takes to map data correctly between systems
  • Treating native tools (like HubSpot's built in AI features) as a full replacement for custom integration work
  • Not budgeting for ongoing sync maintenance once systems are connected
  • Assuming one AI model or one AI tool is interchangeable with another

Not All AI Is the Same

Joe Jerome pushed back on the idea that "AI" is one single thing with one consistent quality level. "I think some AIs are good, depends what you're using, right? And we don't know, when you're dealing with Claude, you can choose the model, you can connect it, you can give it context. Sometimes though, AIs, we don't know what model someone has in their software, we don't know how recently it's been updated," he said.

His conclusion: "The quality of your AI experience can run the range." Comprehensiveness, not novelty, is what separates a useful AI and CRM setup from a gimmick.

Native AI Tools Have Limits

Both hosts discussed the limits of native platform AI features like HubSpot's Breeze. Sajeel Qureshi shared a direct example: "I asked Breeze to do that what you just said and it just bluntly said I can't do that and I won't do that."

Joe Jerome explained why that happens structurally, not just as a product gap: "You're gonna have limits with Breeze, right? Like it's just like how HubSpot can't do it all, and you have that app marketplace. The limit HubSpot has is they have limits in their licensing agreements, their security, whatever else, like has to be commoditized. So they're always gonna be a little limited on what they can do with Breeze."

Sajeel Qureshi agreed and framed it as intentional platform design rather than a flaw: "Breeze is never meant to handle those sorts of edge cases. Just like HubSpot themselves will create native integrations, but not integrations for everything, right? They rely on partners to do custom integrations and handle edge cases to get people onto HubSpot so they can grow better."

The Bigger Shift: Comprehensiveness Wins

Joe Jerome summarized where he believes the real competitive advantage is headed: "I think the real power in CRMs, the real power in AI, the real power in any of this stuff going forward is gonna be in who can do the most comprehensive job. To get the most out of AI and your CRM is gonna be about comprehensiveness. It's gonna be about, are you able to develop a custom AI app? Are you able to integrate all data sources? Are you able to put a process in?"

He added that most companies and even most HubSpot partners are not set up to do this end to end: "I know most partners and most companies cannot do this. They cannot do the end to end solution. And this is where we want to help people is you come in, you figure out your strategy, you figure out what you want it to do. You tell us what it should do and we will take that burden off your hands and make it happen."

Thoughtful AI Versus Lazy AI

Sajeel Qureshi drew a sharp line between AI used carelessly and AI used with intention, comparing careless use to "empty calories." He said: "What we're trying to push onto agencies is thoughtful uses of AI, right? Getting workflows automated, which can be automated, like logging CRMs regularly automatically."

He was equally direct about the failure mode: content or emails that are "mailed it in" with a couple of quick prompts, calling it "a lot of empty calories, nothing in it really. Doesn't taste really good, doesn't read really good. You're just reading, you're just putting it out there for the sake of putting it out there."

Signs of thoughtless AI use

  • Using AI to fully replace human customer conversations instead of supporting them
  • Sending AI generated prospecting emails with no editing or personalization
  • Deploying an AI avatar or chatbot in place of a real person for relationship building
  • Using generic prompts for blog posts, LinkedIn content, or emails without review

Signs of thoughtful AI use

  • Automating repetitive documentation (call logging, CRM updates, task tracking)
  • Using AI to prep meetings so time is spent on decisions, not status updates
  • Keeping a human in the loop to edit and finalize AI drafted communication
  • Choosing a capable, well connected AI model instead of a generic one

What Is Coming Next

Joe Jerome previewed additional tools Computan has built or is building specifically for HubSpot, including an audit tool, a "Magic Module Maker," a code review tool, and a more powerful CRM cleanup tool. He described the scale of that upcoming tool this way: "You can bulk edit ten thousand records at a time. If somebody leaves, you can reassign all of its contacts to the person who should take over. And it will update that field with notes that the new person should have and tasks."

He also described a broader ambition beyond single tools: "We're ready to deploy not only an agency OS, but a marketer OS. Like an operating system basically for this, where we combine our team with a series of prompts we've created," covering morning briefs, to do lists, proposal creation, and case study creation.

FAQ

What does it mean to connect AI to a CRM?

It means using an AI tool, such as Claude, alongside a CRM like HubSpot so the AI can read emails, calls, and chat activity, then draft follow ups, log activity, and prep meetings automatically instead of relying on manual data entry.

Does AI replace the need for a CRM?

No. Joe Jerome described AI and CRM as working together, not as a replacement. The CRM remains the system of record while AI handles the documentation and follow up work that used to get skipped.

Can HubSpot's built in AI, like Breeze, do this on its own?

Native AI features are useful for common use cases but have structural limits tied to licensing, security, and commoditization. Both hosts agreed that deeper, more custom workflows require additional tools and custom integration work beyond what native AI features are built to handle.

Why do agencies struggle to build this kind of AI and CRM integration themselves?

According to Sajeel Qureshi, the difficulty is in doing the data mapping and sync work correctly, not just connecting two tools. Time, budget, and technical depth are usually underestimated.

What tools does Computan use for this kind of workflow?

Their core stack is HubSpot, Claude, Notion, and N8N, with Fireflies used for call transcription that feeds the AI layer.

What is the difference between thoughtful and thoughtless use of AI?

Thoughtful use automates repetitive documentation and prep work while keeping a human involved in final review. Thoughtless use skips human review entirely, producing generic content or replacing real relationship building with an AI avatar or chatbot.