Guide
Context is king – why all customer data needs to live in one place
AI in the CRM is only as good as what it can read. How to gather emails, calls, meetings and contracts in one place, and what AI agents can do with it.
By Viktor Olsson, founder and RevOps consultant
Guide · AI· Updated · 7 min read
Everyone wants to use AI in sales. Few have the thing that makes AI useful: context. An AI assistant that's asked to summarise a customer, prepare a meeting or write a handover can only use what's where it can read it. If half the customer relationship lives in personal inboxes and Slack threads, the AI only sees half the customer. It still answers with the same confidence.
That's why this is the first question to ask for anyone who wants to use AI in sales, marketing and support: is everything about the customer in one place?
What context means in practice
Context is everything that tells you something about the customer:
- emails to and from sales, support and customer success
- calls, recorded and transcribed
- meetings, with notes and transcripts
- SMS and chat
- contracts: start date, end date, notice period, price and terms
- renewals and previous deals
- support tickets and how they were resolved
- notes from everyone who has talked to the customer
Most of this already exists somewhere in the company. The problem is that it exists in ten places, and only the person who was there knows where.
The customer's feature requests to the product team belong here too. We describe how they can be sent from the company record to Linear, with the customer and revenue attached, in HubSpot and Linear.
It matters even without AI
Context in one place is good for the company whether or not you use AI:
- A new rep can take over an account in a day instead of a month.
- Support knows who they're talking to and what the customer has bought.
- Management can trust the forecast, because it's built on what has happened.
- Nothing disappears when someone leaves. Today, years of customer relationships often walk out the door with a departing rep, because everything was in their inbox.
AI just makes the need more obvious. A person can ask a colleague when something is missing. An AI agent can't.
What AI can do when the context is there
Write the handover
When a deal is won, a renewal for example, an AI can read the whole history and write a handover to customer success. That means emails, meeting notes, calls and the quote. Here's what it can look like for the example customer Kvistbro Teknik AB:
Handover: Kvistbro Teknik AB
Renewed: The Pro package, 40 licences. The new contract runs until 31 December 2027.
Key people: Petra Lind, CFO, makes the decisions. Daniel Sund, IT, is the day-to-day contact.
Promised in the negotiation: Fortnox integration live within 60 days.
Risks: Petra mentioned in an email during the negotiation that a competitor has been in touch. Daniel hasn't logged in since August.
No rep writes a handover that good on a Friday afternoon. An AI does it in seconds, if everything it needs is on the customer record.
Find what people miss
Nobody reads a 40-message email thread carefully. An AI does. It can flag:
- that the customer has asked about an API three times without an answer
- that someone promised to get back on Friday, and nobody did
- that the tone has become shorter over the last few weeks
- that a competitor, a budget review or a cancellation is mentioned
An AI step in a workflow can do this automatically, for example when a ticket is closed or a week before a renewal meeting. If it finds something, it creates a task for whoever owns the customer.
Prepare meetings
Before a renewal meeting, the AI can summarise the past year: which tickets the customer has had, which contacts have left, what's unresolved and what you promised last time. It takes a customer success manager half an hour to do that themselves, if they can find everything.
Keep the CRM up to date
Reps rarely fill in fields after a call. An AI can read the transcript and suggest updates: next activity, decision date, which competitors were mentioned. The rep approves with one click. The data gets better, which makes the AI better next time. In HubSpot the feature is called Deal Progression and has been in beta since autumn 2026.
But only on the right foundation
Go back to the handover above. If the email about the competitor had stayed in the rep's personal inbox, the risks line would have been empty. The handover would have looked just as complete, but it would have missed the biggest risk.
That's the risk with AI on an incomplete foundation. The answers look equally good whether or not the underlying data is right, so nobody notices what's missing. More on what it takes in AI in HubSpot.
HubSpot is now built on the same idea
In September 2026 HubSpot launched its autumn release, and it was almost entirely about context. Three parts are worth knowing:
- A CRM that updates itself. Calls, emails and meetings are captured and linked to the right customer automatically, so reps don't have to log them by hand.
- Context Home. A new view that shows how complete your shared context is and where the gaps are, before you let the AI agents loose.
- Agents that read the same context. HubSpot's own example is a sales agent reaching out to a customer the same week a support agent is handling a complaint from that company. Neither knows about the other. That's why the agents in Agent Hub work on the same data.
The features help, but they don't do the work for you. Context Home can show that the reps' emails are missing. It can't connect their inboxes. You do that with the steps below.
How to gather the context
- Connect everyone's inbox and calendar to the CRM and decide which emails get logged. Emails to and from customers should be in; internal emails usually not.
- Call through the CRM or a connected phone system, so calls are recorded, transcribed and land on the right contact.
- Save meeting notes and transcripts on the customer. A private document only helps the person who wrote it.
- Store the contract's key details as fields. Start date, end date, notice period and value. The contract itself can be attached.
- Move decisions from Slack to the customer. The discussion can happen in Slack, but the decision should be written down where everyone can find it.
- Decide who owns data quality. Without an owner, it's nobody's job. In HubSpot the owner can follow the gaps in Context Home every month. What that review can look like is in How to clean up HubSpot before you switch on the AI agents.
Two things to keep in mind: anyone you record a call with should know about it, and personal data in the CRM falls under GDPR. Decide what you store, why and for how long.
Start with one handover
Pick the handover where the most information gets lost today. For most companies that's sales to customer success. Gather everything it needs on the customer, build the handover and measure how often the customer has to repeat themselves. For what the record and the handovers can look like, read One customer, one place.
If you're not sure the problem is bigger than one handover, run the tests in Do we need RevOps?.
Common questions
What does context is king mean? That AI, and people, can only make good decisions about a customer when the whole history is in one place: emails, calls, meetings, contracts, tickets and notes. An AI model can never use what it doesn't have access to.
What is Context Home in HubSpot? A view HubSpot launched in September 2026. It shows how complete your data and context are and where the gaps are, so you can fix them before the AI agents start working on the data.
Do we have to record calls? No, but calls are often where the customer says what never ends up in an email. If you record, the person you're talking to should know, and the recordings fall under GDPR. Decide what you store and for how long.
About the author
Viktor Olsson, founder and RevOps consultant
Viktor has worked with HubSpot for six years, with teams from 3 people to organisations with 30 sales reps across the Nordics. He also works in RevOps at an international SaaS company.