30 Practical Ideas for Sales Managers to Maintain Cognitive Sovereignty While Using AI
Your AI tools are getting better at predicting which deals will close. But this same improvement is silencing the instinct that told you a customer was ready to buy before the data showed it. You are being optimised for what works on average across your team while losing the ability to read what works for your specific accounts.
These are suggestions. Take what fits, leave the rest.
Compare your forecast to Einstein's forecast every monthbeginner
Write down your prediction for your top 20 deals before opening Salesforce Einstein. Then look at what Einstein says. Track where you diverge and why over three months. This keeps your read on deals from becoming passive.
Block 90 minutes each month to review deals Einstein marks as low probabilitybeginner
Einstein will flag deals where the buying committee is stalled or the champion has gone quiet. Read through three to five of these deals yourself. Ask your reps what they see in the customer's behaviour that the data does not show. Write down what you find.
Ask your reps to tell you their read on a deal before showing them the probability scorebeginner
In your one-on-ones, say the account name. Have them describe what they think will happen and why. Only then check what Einstein predicts. Notice which of your reps are reading their accounts well and which ones are starting to trust the score more than their own ears.
Keep a separate forecast for accounts where your instinct conflicts with the modelintermediate
When you think a deal will close but Einstein says otherwise, track it separately. Note the specific signals you are seeing that the model is not picking up. At quarter-end, measure your accuracy against Einstein's on these accounts.
Spend your forecast calls on deals where your read differs from the dataintermediate
Use your forecast review with leadership to argue for deals you believe in but that show as low probability. Describe what you are hearing from the customer. This forces you to articulate your judgement rather than just passing through what the system says.
Ask your reps to describe the buying committee's concerns in their own wordsbeginner
Have them tell you what they hear from each stakeholder when they call. Do not let them read from Salesforce notes first. Then check if what they told you matches what they wrote down. Reps who are paying attention will tell you more in conversation than they recorded.
Review one deal per week where Einstein's probability shifted more than 20 pointsbeginner
Call your rep and ask them what changed. Did the customer actually change direction, or did Einstein's model recalibrate based on activity data. Know which deals you are moving because your team is moving them and which are being moved by the algorithm.
Create a portfolio of accounts where you make a manual forecast adjustment each quarterintermediate
Pick three to five of your largest or most strategic accounts. Forecast them yourself without looking at Einstein first. This forces you to stay skilled at reading deals, not just managing probability scores.
Track which of your reps are closest to your pipeline readintermediate
Over three months, compare each rep's deal assessments to yours. The ones who consistently see what you see are reading their accounts. The ones who defer to the system score are losing their judgement. Invest in both groups differently.
Ask your team what signals Einstein is missing in your top five accountsintermediate
Bring three reps together and show them which of their deals Einstein rates lower than you do. Ask what the system cannot see. Capture those signals. This becomes the basis for the judgement your team keeps and the system cannot replace.
Maintaining Relationship Quality in Outreach
Audit outreach campaign volume before automating sequences in Outreach AIbeginner
Check how many touches your reps are doing manually each week across your top 30 accounts. If Outreach AI scales this by 4x, your reps will not be able to personalise or respond to replies quickly. Set a volume ceiling that keeps time for real conversation.
Have each rep show you their three most recent outreach sequences before they automate thembeginner
Ask them to describe what they are trying to accomplish with each touch and why they wrote the message that way. This is where you spot when someone has become a sequence operator instead of a salesperson. Redirect them before automation scales a weak approach.
Measure reply rate by rep, not just by campaignbeginner
Outreach AI will tell you which sequences get the highest open rate. But some of your reps will get replies from accounts that do not reply to anyone else because they built relationships first. Track who gets replies and make that the goal, not volume.
Review the deals won from automated outreach versus deals won from personal outreachintermediate
Segment your closed deals by source. How many came from a Outreach AI sequence versus came from a rep who called someone personally. You will see that complex deals still come from personal contact, not automation.
Require your reps to write their own first outreach message before they use a templateintermediate
In your one-on-ones, have them write one cold outreach email to a new account by hand. Then show them what Outreach AI suggested. The distance between their version and the templated version shows you how much of their voice they have already lost.
Set a rule that the first touch on a new account must be written by the rep, not automatedintermediate
A customer never hears from your company for the first time through a fully automated sequence. That touch should reflect your rep's thinking about why they are calling this customer. Automation is for follow-up, not for opening doors.
Track which reps get meetings from their outreach and which get their meetings another wayintermediate
Some of your reps will book meetings because they send sequences. Some will book meetings because they have relationships. Notice who relies on which. The ones who rely entirely on sequences will lose their ability to open accounts when the automation stops working.
Listen to recordings of rep calls that came from automated outreachintermediate
Use Gong to review five calls that started from an Outreach AI sequence. Listen to how the rep handles the opening. Are they consultative or transactional. Do they ask questions or deliver a pitch.
Compare the close rate of deals where the rep did relationship building versus deals from pure outreachadvanced
Pull your data and segment by the pattern of activity. Where your rep did three to five personal calls before any sequences, what is the win rate. Where they sent sequences immediately, what is the win rate. The difference is what automation is costing you.
Create a list of accounts where outreach automation is not allowedintermediate
These are your most strategic accounts or accounts where the customer has a difficult buying process. Your best reps work these manually. This preserves your team's skill at relationship selling and shows which reps can still do it.
Keeping Sales Coaching Specific to the Rep
Compare Gong's recommended coaching points to what your rep actually needsbeginner
Gong will flag that your rep should ask more discovery questions. But some reps ask too many questions and kill momentum. Listen to their actual calls. What does this specific rep need to improve.
Coach each rep on what they do differently from the team averageintermediate
If Gong shows that high-performers use more filler words than low-performers, that data is useless. You need to coach to what your specific rep is doing that loses deals. Listen to three calls and name the pattern only you can hear.
Ask your rep what they think they did wrong before you tell thembeginner
After a lost deal, play the call together. Ask them where they think it went off the rails. Some reps will have clear awareness. Some will not. You coach differently depending on their self-awareness.
Coach the behaviours that your best rep uses with your worst rep's accountsadvanced
If your top rep can close in difficult accounts but your struggling rep cannot, listen to how they handle the same customer objections differently. That specific difference is your coaching point. Do not coach to an AI average.
Use ChatGPT to generate coaching ideas, then discard them and coach what you actually hearintermediate
If you are tempted to use ChatGPT to write coaching notes, stop. Ask it for ideas to spark your own thinking. Then listen to the call yourself. What you hear will be different from what ChatGPT generated, and it will be right for that person.
Tell your reps which of their behaviours Gong wants them to change and ask if they agreebeginner
When Gong flags that a rep should change their approach, say it out loud in your coaching call. Ask the rep if they think that is actually the problem. Some Gong recommendations will be noise. Your rep will know.
Identify the specific customer type that each rep struggles withintermediate
One rep might close with CFOs but get stuck with CTOs. Another might close deals fast but fail with slow-moving committees. Name what you see. This is more useful coaching than any AI-generated insight about average high-performers.
Coach your reps based on your own experience, not Gong recommendationsintermediate
Before you watch a Gong coaching video, think about what you have seen work and what you have seen fail. Coach from that experience. Gong is data. Your judgement is what will actually improve this person.
Track which coaching points each rep actually changes behaviour onintermediate
After you coach something, listen to their next three calls. Did they change. Some reps will adapt quickly. Some will not. Notice which feedback sticks and which does not. Adjust your approach to each person.
Coach the rep on what they are not doing, not just what they are doing wrongadvanced
Gong will tell you that a rep interrupted the customer. But the real insight might be that the rep is not building enough rapport before jumping into solutions. Coach towards a positive behaviour, not away from a negative one.
Five things worth remembering
Do not let forecast accuracy improve without asking what your own forecast accuracy is becoming. If Einstein gets better while your instinct stays the same, you have atrophied.
Your best reps are the ones who work in ways that do not scale. Protect some accounts where automation is forbidden. This is where you keep skill alive.
Gong recordings are a tool for you to practise hearing what customers actually care about. Do not use them as a checklist of what your reps should do differently.
The deals that feel hard to close are the ones where your judgement matters most. Do not automate them. Your reps need to stay good at hard things.
Ask yourself each month: what would I know about my pipeline if I turned off all my AI tools for a day and just read my notes and called my reps. That gap is the cognitive sovereignty you are losing.