What sales blogs wrote, October 27. Start AI in sales with a 30-day pilot

I went through what the sales blogs published over the last six days. The AI talk has moved from "should we buy it" to "how do we test it and keep control". I added what I think about each piece and what a dealership can take from it.

Main in this issue#

Start AI with two tasks and a one-month pilot, and let the rep approve every message#

Diego Mangabeira wrote on the HubSpot blog about six tasks you can give to AI. The page was updated on October 23. It is vendor content and it describes HubSpot tools. But the scheme fits any team.

The six tasks are prospecting, lead qualification and routing, and a personal first email. The other three are meeting prep with a call summary, a follow-up after the meeting, and pipeline review. In each one the rep approves every outbound message and every CRM change. The forecast and the decision to walk away from a deal stay with the manager. The author cites a 2024 HubSpot survey in which 64% of sales professionals save one to five hours a week.

The 30-day pilot plan starts with two tasks, lead qualification and the follow-up after a meeting. Week 1 is for a pilot owner and four weeks of baseline numbers. Week 2 is for training reps and a log of edits and rejections.

In week 3 the manager samples the output. In week 4 the team compares results with the baseline and decides to stop, iterate or scale. Stop if reps have to rewrite the output from scratch.

My opinion. The most useful part is the log of edits. If a rep rewrites three emails out of four, the AI does not work, and you see it in a week, not in a quarter.

In a dealership I would take the same two tasks. The first is sorting a new lead and passing it to a manager. The second is the message to a customer after a visit or a test drive. Before the start I would write down how many minutes a reply takes and how many leads get no call. AI prepares, a human decides and answers for it.

Source HubSpot Sales Blog

Before you turn on AI agents, estimate their usage and give access to one team#

On October 22 Nora Pantfoerder, a product marketing manager at Outreach, described how to roll out AI agents so you do not waste usage-based credits. It is vendor content with demo buttons all over the page. The steps do not depend on the product.

The order is simple. Before launch each agent needs an action, a workflow and a team. Then estimate how often and by whom it will be used, and start with one narrow case.

Give access to one team and compare real usage with the estimate. Once a month check who spends what and on which action. Expand when usage is predictable.

The article has an example. Ten account executives use a deal agent five times a week, which is about 200 uses a month. Outreach says the company SailPoint got open rates more than 50% higher with agentic AI. The article gives no baseline, period or method.

My opinion. Agents are often billed by use, and the bill grows quietly. A month later nobody remembers who started what.

In a dealership I would keep a simple table. Which agent, who gave it access, what it does without a human, how many times it ran in a month. No record of each action, no sales team. I would compare the spend with the number of leads that AI brought to a visit.

Source Outreach Blog

Use your own customer base before a bought one, and a lead can book in 10 minutes#

Jason Lemkin, founder of SaaStr, broke down episode 17 of The Agents podcast about growth led by agents. The page shows no date, and the embedded posts are from October 23 and 24. Lemkin gives all the numbers himself, and nobody checked them independently.

For inbound leads the agent looks at first-party data first. That means event attendance, newsletter status, open deals and earlier downloads. Third-party data comes second.

The agent builds a page for the lead and writes an email with a booking link. In the example of Base44 a meeting was booked in about 10 minutes. Lemkin advises starting with one queue and adding steps. Inbound first, outbound around week four, then renewals and support.

On outbound email he is frank. He rates the AI emails at about 5 out of 10 and human emails at 2–3. SaaStr has 3 people and 21 agents. He calls lookalike and competitor lists, built from his own customer list, the strongest outbound play.

My opinion. Every dealership has its own data, and few use it. That means purchases, service visits, the end date of a loan or lease, and earlier leads.

Before each call to the database I would give the manager a short brief of what the customer already did with us. AI builds the brief, a person makes the call. A score of 5 out of 10 is a reminder that an AI email is not ready without a manager's edit.

Source SaaStr

Buyers already ask AI whom to choose, so check what it says about you#

In another SaaStr post Lemkin claims that software buying decisions are shifting to AI agents. The page shows no date, and the embedded posts are from October 16 to 20. SaaStr runs more than 21 agents with three people. Lemkin tries a new tool only if his "AI VP of Revenue" agrees.

The examples from the post. An agent picked the host Render, and later the agent moved him from SendGrid to Resend. By his account agent traffic on Vercel went from under 3% of deployments to more than half in a year. An agent tested a vendor's pricing and dropped it in under 12 hours.

Lemkin's advice is this. Find about 10 people with agents in production in your category.

Every two weeks ask them to query the agent about which tool to use for your job, and record the answer. Find the phrase the agent uses to rule you out. Treat documentation and the interface for software as sales assets. Make pricing that survives a calculation.

My opinion. For a dealership this is no longer about software. A customer asks ChatGPT where to buy a model in my city and with what loan.

Today I would ask five such questions and write down whether our dealership is named, what is said about it and where that comes from. Then I would fix the site. Prices, terms, addresses, reviews, delivery times. An agent takes facts it can verify, and nice words do not help.

Source SaaStr

A great seller makes buying easier, and a weak one makes it harder#

Lemkin answered a reader question about the difference between a good sales exec and a great one. The page shows no date. The blog list shows October 22, so the text may have been published earlier.

He says a great rep often closes about 9 times more than a poor rep. The rep also beats a mid-pack rep by more than 2.5 times, on the same leads. The gap has three parts. Great reps close bigger deals, discount less and close faster.

Great reps listen to the customer instead of following a script, know the product and the competitors, and are honest about weak spots. They create urgency with new value, such as demos, help for other stakeholders and trials, not with a pile of emails.

Two more numbers. A great rep spends 10 times more time preparing for meetings and calls. A weak rep sells to whoever seems to lead the project, and a great rep gets buy-in from everyone who matters. For the age of AI Lemkin adds a point. A rep must know the product better than AI does, or the buyer has no reason to talk to a human.

My opinion. From this list AI covers preparation, not knowledge. It builds a customer brief in a minute.

I would give managers one rule. Prepare for each call from a brief. At the team meeting, review one recording and ask if it got easier for the customer to buy. Ask for a discount only in exchange for something.

Source SaaStr

Also published#

• Salesforce Blog. Beyond Collaboration. CRM and Slack Are Slaying Sales Goals, October 26, deals and approvals in Slack, promotes Salesforce products.
• HubSpot Sales Blog. RevOps automation. What it is and how it works, updated October 26, six steps to roll out revenue automation.
• Outreach Blog. AI Agent Interoperability. How Connected Agents Align Sales and Marketing, October 22, shared definitions of a qualified lead first, agents second.
• SaaStr. Should Your VP Sales Start Off as a Player-Coach?, the page has comments from 2013, so it looks like a re-run.
• SaaStr. Does Your CRO / VP of Sales Need to be An Expert In Your Product?, a new leader must master the product in 30–45 days.
• SaaStr. Dear SaaStr. What's The #1 Most Important Step in Closing a Deal?, a fast and thoughtful reply, the page shows traces of 2022.
• SaaStr. The Double Hockey Stick. Why 1% of Your Users Now Drive 75% of Your AI Costs, how Figma, Clay and Higgsfield limit costs.
• SaaStr. Dear SaaStr. As a Co-Founder, When Do You Know It's Time to Quit?, Lemkin advises a pause and a talk with a mentor first.
• SaaStr. Dear SaaStr. How Do You Know When It's Time to Sell Your Startup?, companies are usually bought, not sold.
• SaaStr. 20VC x SaaStr. $20M ARR and Unfundable, Anthropic's $16.8B Quarter, and Modal at $15.75B, a weekly venture roundup.
• SaaStr. Getty Images Files for Chapter 11 at a ~$910M Run-Rate, a bankruptcy filing at about $910 million in revenue.

The Salesloft Blog did not open. The list of posts did not load on the blog page or at the second address. Apollo.io and Pipedrive opened, but their posts show no dates. The second address returned an error, so I could not confirm any post from this period.

These blogs had no new posts in these days. Gong, LinkedIn Sales Blog, Close, Sales Gravy, RAIN Group, Harvard Business Review, The Bridge Group and The Sales Blog by Anthony Iannarino. From Salesforce I took only the post about sales. The other five feed items in the period were about other topics.