whynotroman

News · Oct 22, 2026 · 11 min

What sales blogs wrote, October 22. AI agents in sales need trust and control

Outreach on rep trust in AI agents and on proving ROI, SaaStr on two agents with a shared mailbox, RAIN Group on weak sales managers, Sales Gravy on December without discounts. Short facts, my opinion, a link to the article.

I went through what the sales blogs published over the last three days. Most of them share one theme. Teams already launch AI agents, but rep trust, ROI proof and human control still fall to the sales leader. I added what I think about each piece and what a dealership can take from it.

Main in this issue

Start AI with call review, not with outreach, and reps will trust it

On October 21 Josh Harmon, a go-to-market leader at Resi, wrote on the Outreach blog about how his team rolled out AI. They did not start with automatic outreach to customers. They started with call coaching. He says an AI coach is objective and can be both critic and cheerleader, while reps tend to dwell on their bad calls.

In 2022 Resi introduced call recording and presented it as "a coach in your pocket". A manager cannot join every call. Agents take over the prep work, which means account briefs, research and a first draft of the message. Harmon does not want a rep to become a data analyst for two hours a day.

He suggests asking reps not whether they like the agent, but what the agent did for them. He compares guardrails to bowling bumpers and says reps should help design them. The article has no results data. It is one leader's view on a vendor blog.

My opinion. A rep fears being watched, not AI. If the first AI in the team only checks their calls, they resist it. If it helps them prepare and explains their mistakes, they ask for more.

In a dealership I would start with call and chat review. After a week I would ask each rep what the agent did for them personally. I would launch outreach in the rep's name later, when the team already trusts the system.

Source Outreach Blog

Record your own numbers for 60–90 days before you buy AI, or you cannot prove ROI

On October 21 Anthony Zumpano described on the Outreach blog how to build a business case for AI sales tools. It is vendor content and it mentions Outreach products. But the math is general, and any team can use it.

Count the full cost. That means the license, CRM integration, cleaning up the database, training, and manager time to review results. He quotes a Gartner forecast that 60% of AI projects will be abandoned by 2026 without a ready database. A McKinsey rule of thumb in the article says about $3 of change management for every $1 of model development.

Discount the benefits. An Outreach Insights Group 2026 report says AI saves 7–8 hours a week, but not all of that time goes into selling. The author applies a capture rate of 25%, 50% or 75%.

He says to state win-rate gains in percentage points, not as a percent of a percent. Count only reps who still use the tool. He gives a case where AI use in Microsoft's sales organization fell from 22% to 5% in a month.

His main advice is to record a baseline over 60–90 days, including a month-end and a quarter-end, and only then buy. Ask finance for a pilot stage with go or no-go thresholds, not the full sum.

My opinion. Dealers often buy AI after a demo, and three months later they cannot say what changed. The reason is simple. Nobody measured anything before the purchase.

I would write down three numbers over two months. How fast a lead gets a first reply, how many leads never get a call, and how many booked visits come to the showroom. Then I would buy a pilot stage against those numbers. If the tool does not move them, you see it at once and not at the end of the year.

Source Outreach Blog

Two AI agents built a shared mailbox on their own, and a human's safe word decides

Jason Lemkin, founder of SaaStr, described two agents. One is Muse, Meta's assistant, used by SaaStr's Chief AI Officer Amelia. The other is 10K, SaaStr's "AI VP of Revenue", built on Replit and connected to more than 20 apps. The page shows no date, but it embeds a post from October 19.

No ready connector linked them. Muse proposed a shared mailbox, a plain endpoint on 10K's server where both post messages. The only human step was issuing an access key. Each morning 10K sends Muse a revenue brief. It lists contracts waiting for approval, overdue invoices and what Amelia should do that day.

A few days later Muse noticed a failing sync that 10K had not reported to a human. Amelia reconnected it. A guardrail was added too. Muse cannot trigger an action in 10K until Amelia types a keyword in her chat with Muse. Lemkin advises tracking every key an agent receives, requiring human approval for actions between agents, and shutting the channel if people cannot read it.

My opinion. The trick with the mailbox is not the point. One agent checked another and found an error, and the decision stayed with a person. That is my scheme. AI prepares, a human decides and answers for it.

A dealership will soon run several agents. One in the CRM, one in the messenger, one in the call center. I would write down in advance who gave each one access and what it may do without a human. No record of each action, no sales team.

Source SaaStr

Pipeline grows but wins do not, because the sales manager does not coach on the job

On October 21 Bethany Talbert wrote on the RAIN Group blog about why behavior change after training does not last. RAIN Group's 2026 research found that 47% of companies saw growth in qualified pipeline. Only 40% saw more wins, and 37% saw higher quota attainment.

She then uses leadership research from Alchemist. 80% of RAIN respondents said developing sales managers is challenging, and 73% said the same about coaching the sales force. Top sellers are 83% more likely to have a strong manager. In Alchemist's data, 72% of leaders find it hard to sustain behavior change. Firms that use one leadership framework are 28% less likely to struggle with accountability.

Her advice is simple. Pick one behavior and describe what it looks like. Set check-ins at 30, 60 and 90 days. Before training, ask what participants will do differently the next day and what their managers will do differently. In one example a CRM was changed so that a field required the new behavior, and adoption rose at once.

My opinion. A strong seller is promoted to manager and never taught to manage. Then everyone is surprised that training is forgotten a month later. It is a familiar picture in any dealership.

I would pick one behavior for the quarter. For example, the question at the first meeting, "Who else will decide with you?" The manager listens for it in calls. Then the team reviews recordings together. The rest can wait. AI helps here because it listens to all calls, not five out of a hundred.

Source RAIN Group Sales Blog

Do not cut prices in December, go back to people who already know you

The Sales Gravy article on December came out on October 21. The page shows no author line, and the bio at the bottom belongs to trainer Jessica Stokes. The text has no numbers. Jeb Blount wrote about leftover customer budget on October 19, and I covered it in the last issue. This piece has twelve other tips.

The main idea is that December discounts teach buyers to wait for year-end. The fastest sales come from people who know you. The author suggests reviving deferred purchases, going through every stalled proposal, and offering add-ons that customers put off. Look also for small decisions, such as one more unit, because those get approved fast.

For large customers she suggests a year-end review. Show the results and ask what they want done before the year ends and what they plan for next year. For the forecast you need the next step, the decision date and the signer. If a buyer asks for a discount, find out what sits behind it, budget, timing or a competing quote. If you give something, ask for something back, such as a signed agreement by a set date, a larger order or a longer term.

The article ends with a plug for Sales Gravy's closing training. The advice on discounts stands on its own, but it is an ad.

My opinion. Dealerships run December on discounts, and buyers know it. If a customer asks for a discount, I would first ask what is behind it. After that, a concession only for something in return. Payment this year, a trade-in on the spot, extra equipment in the order.

One more step for this week. Take people who said "we'll think about it" in spring and summer. Call them with one question about how they plan to start next year. AI can build the list from the CRM, and a salesperson makes the call.

Source Sales Gravy

Also published

The Salesloft Blog did not open, and the list of posts did not load on the blog page. Apollo.io and Pipedrive opened, but their posts show no dates, so I could not confirm any post from this period. Gong, LinkedIn Sales Blog, The Sales Blog by Anthony Iannarino and Harvard Business Review had no new posts in these days. From Salesforce I took only the posts about sales.

Next note · 13 minWhat sales blogs wrote, October 20. AI finds the reason, a person makes the call →HubSpot doubled outbound conversion with an AI agent, SaaStr on how to raise win rates, Kawasaki Engines and its dealer portal, Jeb Blount on year-end budgets. Short facts, my opinion, a link to the article.
#sales blogs#sales#ai

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

Read next

← All notes