AI for Sales Enablement | Week ending October 4, 2026

One thing I’m looking for in AI is the ability to take recurring work off the plate of enablement and the teams we support. Account research. Learning-module development. Keeping the knowledge base aligned with what’s actually happening in the business.

This week’s three picks are worth looking at through that lens: GPT-6.1 Sol for complex tasks, OpenAI dots for work that continues between conversations, and Google skills for reusable instructions.

The examples below are ideas to test. They aren’t results from completed experiments.

OpenAI positions GPT-6.1 Sol as approaching Astra’s performance on complex professional work at lower API rates. For enablement, that makes it a candidate to test on account research, learning modules, and cross-checking a knowledge base.

The pricing detail matters. Standard API rates are $2 per million input tokens and $10 per million output tokens, one-fifth of Astra’s rates. Ordinary input and output prices are unchanged from GPT-6 Sol; cached input drops from $0.20 to $0.10. These are API prices, separate from ChatGPT subscriptions, and don’t guarantee the same reduction in cost per completed task.

A useful first test: give it an approved pricing update and the related FAQ and training material. Ask it to flag contradictions, stale answers, and internal guidance that would be misleading in customer-facing content. Require a source for each finding and a proposed correction for the content owner to review.

The model needs the relevant tools and access. Scheduling comes from the surrounding workflow. Availability is in Work and Codex on eligible plans, plus the API.

Dots are ongoing assistants that can work between conversations using connected apps and a cloud computer. They’re powered by GPT-6 Astra. The potential benefit for enablement is assigning a clearly defined responsibility that continues after the first answer.

Think about launch readiness. A proposed assignment could be to check named product sources, identify changes affecting rep materials, and prepare a short list of updates for review. Another could be reviewing recurring rep questions to find gaps in the knowledge base.

That second example is especially interesting to me. Several versions of the same question might point to an answer that’s missing, hard to find, or written for the wrong situation. I’d want the AI to show the evidence and help distinguish those possibilities.

Manager coaching preparation, onboarding follow-through, report pulling, and data analysis are other candidates. Start with one. Define the sources, timing, output, and approval boundaries. Connections and permissions determine what it can do; adding a messaging channel alone doesn’t start monitoring.

Access is still rolling out, with plan, region, and administrator requirements. Test the quality of its findings and the review time before expanding the assignment.

Google’s September 30 announcement brings reusable skills into Gemini chat. Save instructions for a task and invoke them again, including combining multiple skills in a request.

For enablement, a possible starting point is a launch-brief skill that asks for the change, affected audience, and action reps should take. Pair it with a plain-language skill and a rule to mark missing facts as unknown. That could make a familiar process easier to repeat without rebuilding the prompt each time.

Availability needs a careful read. Personal-account rollout began September 30. For Workspace, Rapid Release is scheduled to start October 5, Scheduled Release October 19, and the Gemini-app rollout for Workspace users October 13. Google’s rollout schedule has the eligibility details.

Maintenance matters here, too. Gemini-app and Workspace skills don’t sync with each other. Instructions developed in Google Docs require a manual Sync with Docs step to update the Studio copy. Someone needs to own keeping the procedure current.

My trust test is whether the AI can verify its sources and bring back insights that aren’t obvious.

For a knowledge-base check, that means showing which answers conflict, explaining where a rep could get into trouble, and identifying who should review the correction. Changes to the authoritative answer still need human approval.

That’s where I’d start: one recurring task, clear evidence, and a review step. Then ask whether it found something useful and actually took work away from the people we support.