whynotroman

News · Nov 05, 2026 · 9 min

What sales blogs wrote, November 5. An agent does the work, it does not only give tips

Salesforce on leads that reach a seller in 20 seconds, HubSpot on finding buyers from CRM signals and on Claude inside the CRM, Outreach on the hours an agent gives back and on call coaching. Short facts, my opinion, a link to the article.

I went through the sales blogs for the last three days. The main idea is simple. AI that only gives tips leaves the work with the seller. AI that does the task gives the seller time back. I added what I think about each piece and what a dealership can take from it.

Main in this issue

A lead reaches a seller in 20 seconds instead of 20 minutes

On November 4, Salesforce authors Kaitlin Lee and Jessica Radloff described how the company rebuilt lead routing. Before, the rules lived in three systems, and almost every inbound lead went to a seller. A demo request looked the same as a downloaded article.

Now everything about a person sits in one profile. Two models give each lead a score from 1 to 5 and name the product the person is most likely to buy.

Then there are three routes. Strong leads go to a seller. Leads with interest but low readiness go to marketing. An agent starts the talk with the rest and hands them to a seller when the buyer is ready.

The authors say the numbers describe the whole project, not one tool. Lead assignment fell from 20 minutes to 20 seconds. A rule change takes days instead of four weeks. Sellers work with about a quarter of the leads. The readiness threshold sits in a table, and an SDR manager raised it for one product without a code release.

My opinion. For a dealership, the point is this. A test drive request and a "what is the price" request should not wait in one queue. I would split them into two routes and measure the time to the first reply from a live manager.

The rule "every lead has a destination" works without big systems. In my measurements at dealers in different countries, 40% of requests come outside working hours. They need their own route, for example an instant reply from an assistant and a call from a manager in the morning.

Source Salesforce Blog

A signal from your CRM is worth more than a bought list

On November 4, Althea Storm of the HubSpot blog explained automated prospecting. She suggests starting with signals the buyer left on their own. These are a filled form, repeat visits to the pricing page, a stage change, and a closed-lost deal that comes back. The author says scraped lists often damage the sender's reputation.

The example points are these. A demo video gives 15, a webinar 10, an enterprise pricing page 20. A contact with 60 points gets the next step. A buyer leaves the sequence if they book a meeting, open a support ticket, or are in a renewal. The author also says to check consent before any message.

The author keeps discovery talks, objection handling, and fine personalization with people. A person reviews every email that AI drafts. On timing, the effect on routing and response time shows in one to two weeks, and meetings grow in two to four. The article has no research, it is a guide.

My opinion. A dealership already has signals. A buyer opened a trim page three times, ran a loan calculation, or looked at service. I would put three or four such events with points in the CRM and create a task for the manager at 60 points.

I also take the suppression rule. A customer whose car is in service, or who has an open complaint, does not get a promotion email. That is a CRM field, not a matter of manners.

Source HubSpot Sales Blog

Claude in the CRM reads records, writes only with approval, and deletes nothing

Rachael Nicholson of the HubSpot blog explained how the HubSpot connector works with Claude. The page was updated on November 4. This is HubSpot's own material about its own product.

Claude reads contacts, deals, companies, tickets, and the history of contacts with a customer. It can create and change contacts, deals, tickets, and activity records. Leads, invoices, orders, and subscriptions stay read-only. The connector does not support delete at all, and the page calls this a guard against losing contact data by accident.

The author says to set write actions to "needs approval". Each change goes to an audit log under both the user and the connector. Only an admin can connect it, and the Claude account needs a paid plan. With sensitive data on, Claude cannot see activity records.

My opinion. This shows the order I would use to let AI into a dealership CRM.

First reading and summaries. Then email drafts. Then writing with a manager's approval. Delete for nobody.

My first request would be this. "Show deals with no contact for 14 days and suggest the next step for each." The manager decides what to send. AI prepares, a person decides and answers.

Source HubSpot Sales Blog

An agent that only advises leaves the work with the seller

On November 3, Maria Akhter of the Outreach blog wrote that read-only AI is limited by headcount. A tip adds a task, the seller's hours do not grow, and revenue grows only with hiring. This is vendor material, and it also promotes the vendor's agents.

The author uses outside and own sources. By the Salesforce report, a seller spends 40% of the week selling. In Outreach's own report, meeting prep fell from 60 to 23 minutes. Active AI users save 15 to 21 minutes a day on CRM entries and call summaries.

The author's math goes like this. Eight meetings a week give back about five hours, and entries and summaries add about 1.5 more. Ten sellers who put five hours into calls give 50 hours, which is about three sellers. Count only hours that became talks with customers. A person should approve changes to stage, amount, and close date, or the forecast gets worse.

My opinion. I would test this with one manager. For a month I would count the hours a week that go to prep and entries, and how many came back as calls.

The returned time should go into the calendar as calls to the customer base, or it will disappear. Before a request for a new manager, I would compare these hours with the cost and ramp time of a hire.

Source Outreach Blog

One scorecard for everyone, then AI scores every call

On November 3, Maria Akhter of the Outreach blog wrote about call coaching when a team grows. No manager can listen to all calls and coach everyone the same way. The author compares four coaching models, among them GROW and FUEL. This is vendor material too.

The plan has five steps. First, choose one model for all regions. Then write a scorecard of three to five visible behaviors, so two managers score one call within one point.

Next, AI scores every call. At first it only flags moments, and live tips come later. Manager time goes to the flagged moments. The author suggests measuring the change in behavior from a baseline, not by logins.

The example team has 75 sellers in three regions. The author says to start with one region and write the scorecard before AI scores the first call. Trust, emotion, and big strategic talks stay with the manager.

My opinion. I would write the scorecard for a dealership myself, on paper. Three points. The manager learned what the customer needs the car for, named the next step, and agreed on a visit date.

Then I would let AI listen to a week of calls and show the ones where points are missing. The sales manager reviews those. Show the scorecard to managers first. AI does not replace the manager, it shows who did not do the work.

Source Outreach Blog

Also published

Salesloft Blog did not open. The list of posts did not load on the blog page or in the feed. The Apollo.io, Pipedrive, and LinkedIn Sales Blog pages opened but showed no dates for the period, so I could not confirm new posts there.

There are no new posts in six blogs. These are Gong, Close, RAIN Group, Harvard Business Review, The Bridge Group, and The Sales Blog by Anthony Iannarino.

Next note · 3 minSales blogs, November 17. How to check your team's results →Four reads on marketing targets, deal stages, sales metrics and AI value. The main ideas and practical checks for a sales team.

Roman Spitsyn. I build sales teams where AI agents do the work and people close the deals.

Read next

← All notes