An AI sales assistant that writes a nice email is cute. An account executive at an IT MSP needs one that does the admin.
Here’s why. An MSP account executive’s week is rarely one kind of deal. There’s a QBR (quarterly business review) to prep, a managed services renewal coming due, and a project upsell waiting on a next step. Then a discovery call with a new prospect lands in the middle of it.
Every one of those leaves admin behind. Notes to log, a stage to move, a close date to fix, and a follow-up to write while the call is still fresh.
Most tools sold as AI sales assistants summarize the call, suggest a next step, and hand the work back to you. That’s a to-do list with better grammar.
I built Go Fig to do the work instead. Go Fig is the sales admin for account executives at IT managed service providers (MSPs). It helps an AE stay on top of their pipeline and move deals forward in HubSpot, Salesforce, and all other major CRMs, and it does not replace the PSA.
The full picture for MSP sales teams lives on our IT MSP page. This post goes deep on one question: what should an AI sales assistant do for an MSP account executive, and what should it never do?
What should an AI sales assistant do for an MSP account executive?
It should take the admin off your plate and leave the selling to you. In practice that’s five jobs, and a good assistant does all five in the CRM and inbox you already use.
- Hand you the day before you open your inbox. A brief that says who is waiting on you, which deals went quiet, and what’s on your calendar.
- Log the call where the deal lives. After a discovery call or a QBR, the note, the next step, and the deal update land on the right record in your CRM.
- Draft the follow-up while the call is fresh. The draft covers what was said and what was promised, written in your voice with the deal history in it.
- Catch what slipped. Deals with no next step, close dates that already passed, and deals missing an amount get flagged while they’re still easy to fix.
- Get the handoff ready. When a deal closes, the kickoff email, onboarding plan, and account notes are ready for the team that sets up the new client.
Notice what’s not on the list. Telling you what to do. A suggestion you still have to act on is not admin taken off your plate.
Here’s how those five jobs map to a normal week:
| Job | When it runs | What you do |
|---|---|---|
| Daily brief | On a schedule, like 7 AM every weekday | Read it before your inbox |
| Call logging | After a discovery call or a QBR | Approve the note and the deal update |
| Follow-up draft | The same day as the call | Approve it, edit it, or skip it |
| Pipeline check | Every night | Approve the fix or handle the flag |
| Onboarding handoff | When the deal closes | Review the kickoff email and plan |
Call logging, follow-ups, and CRM hygiene each deserve a deep dive of their own. Here I want to stay on the parts that decide whether an assistant helps you or hurts you.
Why should it do the work instead of suggesting it?
Because a suggestion is still your work. Picture two assistants after the same QBR.
Assistant A sends you a summary. Three bullet points and a recommendation: “Follow up on the firewall refresh.” You still open the CRM, find the deal, update the stage, fix the close date, and write the email. The summary saved you the part you were going to remember anyway.
Assistant B has already drafted the note on the right record. It has staged the stage change and the new close date. It has written the follow-up in your voice with the refresh timeline you promised. All three wait for your OK. You read them, fix one line, and approve.
Assistant A is a to-do list. Assistant B is a sales admin.
The difference shows up after every call. At an MSP the calls don’t stop, because the account doesn’t close once and go away. It comes back every quarter with a QBR, every year with a renewal, and every time a project comes up.
Should an AI sales assistant act without your approval?
Not at first. And for some actions, maybe never.
Your clients trust you with their systems. An assistant that emails a client the wrong renewal terms on its own puts that trust at risk, and no amount of saved time makes up for it. I wrote more about this in Draft, never send, and the same rule holds for MSP sales.
So here’s how approval should work:
- Start with everything waiting for your OK. Every draft, note, and update waits until you approve it, edit it, or skip it.
- Graduate one action at a time. Logging a note can run on its own while client emails still wait for you.
- Take it back anytime. Anything you let run on its own can go back to waiting for your OK whenever you want.
In Go Fig, approval is set per agent and per action. Drafts wait in the Feed, the one list of everything waiting for your OK. You can approve from the Feed, from email, or from Slack, where each draft gets Approve and Reject buttons.
Approved replies go out from your own Gmail or Outlook, not from a shared mailbox. The client hears from you.
Writing back to your tools is its own switch, too. An admin turns it on for the company, then for each tool. Every change Go Fig makes is recorded in a log your team can review.
What should it read before it writes anything?
More than CRM fields. What happened on a deal is in the inbox, on the calendar, and on the call. An assistant that only reads fields writes generic follow-ups, because it only knows generic facts.
Here’s what Go Fig reads:
- Your CRM. In HubSpot, the deals, contacts, and companies, plus the calls, emails, meetings, notes, and tasks logged against them. In Salesforce, the accounts, contacts, opportunities, and the tasks and meetings on them. HubSpot custom properties come through too, and Salesforce custom objects, like renewals or projects, are added on request.
- Your inbox and calendar. Gmail or Outlook, so a draft knows what the client already asked and when you’re meeting next.
- Your call recordings. If your team records calls in Fathom or takes them on Dialpad, Go Fig reads the transcript and drafts the CRM note from it. It only reads. It never starts a recording or places a call.
At an MSP that context matters more than most places. The QBR follow-up has to know which project you said you’d scope. The renewal check-in has to know what the client said about timing on the last call.
You can also ask Celeste, Go Fig’s AI analyst, about your own pipeline in plain English. Ask for “my renewal deals closing this quarter” and you get the answer with the records it came from.
What should an AI sales assistant leave alone?
Three things, and a good vendor will tell you up front.
Your PSA. ConnectWise, Autotask, and HaloPSA run tickets, agreements, and billing. A sales assistant should work the pipeline, not service delivery. Go Fig does not replace your PSA, and there’s no PSA connector today. It doesn’t read from or write to ConnectWise, Autotask, or HaloPSA. Setting up a new client in the PSA after the deal closes is still your team’s step. Here’s how Go Fig fits beside your PSA.
Your quotes. Quoting stays in your quoting tool or PSA. Go Fig does not build quotes. On the Team plan it can draft proposals and statements of work from email, the CRM, or a PDF.
Your lead sourcing. Go Fig works the pipeline you already have, from first touch through close and onboarding. It does not find new leads. That’s a marketing job, and we’re happy to point you to a partner who does it well.
If you want one tool that quotes inside ConnectWise and runs sales and service in one place, Go Fig isn’t it. I’d rather tell you that here than on a demo.
How do you test an AI sales assistant before you buy it?
On your own deals. A demo account full of perfect data tells you nothing about what an assistant does with your real pipeline.
- Connect the real CRM and inbox. You should sign in with your own account and never hand over a password or an API key.
- Pick three recent deals. A renewal, a project upsell, and a discovery call with a new prospect. Each one produces different admin, and you want to see all three.
- Read the drafts like a skeptical sales manager. Is the note accurate? Did the follow-up promise only what you promised? Would you send it with one edit or ten?
- Check what it changed. Every change should have waited for your OK, and every change should show up in a log.
- Ask the security questions early. Ask about SOC 2, where your data is stored, and whether it trains anyone’s model. Go Fig is SOC 2 Type II audited. Data from the apps you connect is stored in Google Cloud in the United States. We only use AI providers’ paid business plans, whose terms say your data isn’t used for training. The details are on our security page.
That’s the whole test. If an assistant passes it on your deals, it’ll pass it on Monday morning.
If you want the short version of what the agents do for an MSP account executive, it’s on the IT MSP page.
See it on your own pipeline
A solutions engineer connects your CRM and inbox with you in a 30-minute demo. You see the brief, the call notes, and the follow-up drafts on your own deals, and you decide what waits for your OK.
