Making AI Work in Your Sales Team
Making AI Work in Your Sales Team
Every quarter I (ask Claude to) review what has changed in the world of sales technology.
This quarter Salesforce, Gong, Salesloft, HubSpot and Zoom all announced AI agents that can forecast, prospect, update records and coach. The technology is improving quickly.
What has not changed is what decides whether any of it works:
– how your team sells,
– whether they use the systems you give them, and
– whether those systems connect to each other.
Here is what I would take from this quarter.
1. Fix the process before you add AI
A Salesloft survey of US revenue leaders this quarter found that every one of them was using AI somewhere in their sales process, but only about one in five had it working in a way that delivered measurable results.
Most teams are adding AI faster than they are putting in the measures that would show whether it is working.
AI speeds up whatever process you already have. If your pipeline stages are not defined, your qualification is inconsistent, or your CRM is half filled in, it will speed that up too.
2. Behaviour decides the result, not the software
Gartner predicts that by 2028 AI agents will outnumber salespeople ten to one, yet fewer than 40% of salespeople admit that agents improved their productivity. The gap is not the technology. It is whether people change how they work.
Every AI sales tool depends on reps recording their activity and on managers using the information. If the sales manager runs the weekly pipeline review from memory and a spreadsheet, the reps will treat the CRM as admin, and the AI will have nothing reliable to work with.
The behaviour that matters most is the manager’s. When the manager runs every review from the system and asks questions based on what it shows, the team follows.
3. Integrate before you add
I see many sales teams with too many separate tools: a CRM, an email system, a call recorder, a quoting tool and several spreadsheets, none of which talk to each other.
Adding an AI tool with its own login usually makes this worse.
The more useful development this year is AI that works inside the systems you already have.
Both Claude and ChatGPT can now connect directly to CRMs such as HubSpot, Zoho and Salesforce.
A sales manager can ask for every open deal over £20,000 with no activity in three weeks, or a one-page briefing on a customer before a visit, from the data already held.
Before buying anything new, ask one question: how does this connect to our CRM and email, and who keeps that connection working?
This matters more now because the large vendors are merging. Salesloft has taken on Clari, and Seismic has acquired Highspot. Salesloft has said that fully combining its products is still years away. Short contracts are sensible until that settles.
4. Make qualification honest
Most pipelines carry deals that should have been closed out months ago. A newer group of tools scores each opportunity against a qualification framework such as MEDDPICC using evidence from calls and emails, rather than the fields a rep chooses to fill in. If the economic buyer has never been on a call, the tool flags it.
Whether you use one of these tools or not, the principle is worth adopting: a deal is qualified when there is evidence, not when someone ticks a box. That single change does more for forecast accuracy than most software.
5. Support your salespeople, don’t replace them
AI “sales development reps” that write and send outreach automatically are heavily marketed. Between half and 70% of customers stop using them within a year.
Teams that use AI to support their people generate about 2.8 times more pipeline than teams that try to replace them.
In a market the size of Northern Ireland, where buyers know each other, a badly targeted automated email gets noticed and remembered.
Use AI for research, preparation and first drafts, and keep a person responsible for what is sent and for every reply. If you do automate outreach, check the UK PECR rules: sole traders and some partnerships are treated as individuals and need consent.
6. Make coaching a weekly habit
AI roleplay and call-scoring tools let a rep practise a pricing conversation or a difficult objection against a simulated buyer, then see where they lost control of the conversation.
Call recording tools flag coaching points from real calls but ultimately none of this replaces a good manager or coach.
What it changes is frequency. Practice can happen every week instead of once a year on a training day, and the manager can see who needs help on what before it shows up in the numbers.
7. Prepare for buyers who use AI too
Buyers increasingly research suppliers with AI before they contact anyone. Gartner found that 69% still go to a salesperson to check what the AI told them.
Two things follow. Make sure the product, specification and pricing information on your website is accurate, because AI tools will repeat it. And prepare your reps to confirm, correct and add detail, rather than repeat the brochure.
That is a different conversation, and most teams have not been trained for it.
If you want to go deeper, get in touch
Thanks for reading
