Trent Gillespie
January 29, 2026
Most business leaders still think "AI" means a chat box in a browser.
Your team copies context from Gmail, pastes it into ChatGPT, gets an answer, then pastes it back into Slack.
That's not transformation. That's faster context switching.
What's changing now is simple and brutal: AI is moving into your tools, pulling context itself, and starting to execute multi-step work. And the gap between businesses that get this and those still treating AI like a fancy search engine? It's widening fast.
Early data from companies deploying AI at scale shows something counterintuitive: the highest-value skill isn't prompt engineering—it's AI supervision. Teams that learned to review, validate, and course-correct AI outputs across multiple systems achieved 2-3× better outcomes than teams still focused on crafting perfect prompts. The work shifted from input to oversight.
The capability gap isn't about better models. It's about businesses that learned to integrate AI into workflows versus those still copy-pasting between browser tabs.
An open-source project called Clawdbot (now Moltbot) dropped this week, gaining a lot of attention. It's a self-hosted personal AI you run on your own computer (or virtual one) that takes action and communicates inside the channels you already live in: Slack, Teams, iMessage, and more.
You don't need to deploy it today—it still requires technical expertise and carries security considerations that aren't worth it for most businesses. But it signals that we are getting closer to always-on assistants that feel local, persistent, and embedded in real workflows, not just a web-browser.
Clawdbot/Moltbot is just the latest entrant in the race to provide agents that can do work on your behalf, where you work:
ChatGPT Agent allows an agent to control a browser by interacting with on-screen elements—clicking, typing, navigating like a person would—plus doing any other task that ChatGPT does, but better.
Anthropic Claude’s Cowork pushes agentic work onto the desktop: install the desktop app (Mac only today), and now Claude can organize, create, edit, and assemble files and complete multi-step tasks.
Microsoft has already announced multiple times they want Windows to become your primary agent—though others have beaten them to demonstrating what that actually looks like.
These aren't production-ready tools today, but it’s coming fast. Soon, telling AI "prepare the monthly board report" and having it navigate your systems autonomously won't be science fiction—it'll be Tuesday.
And if you're still learning "basic ChatGPT prompting" when that future arrives, you've already lost.
Last week I covered how ChatGPT Apps let AI work directly inside your business systems.
This week, the story accelerated. Anthropic released Claude Connectors, matching that capability, providing connectors to Gmail, Notion, and other (non-Microsoft) applications.
The direction is clear: AI will become your primary work interface, not a separate tool you visit.
The businesses still treating AI as "better search" or "writing assistant" are falling behind businesses that learned to supervise AI working across systems. And that gap compounds fast.
According to research from RescueTime, knowledge workers lose 60% of their workday to fragmented work and context switching. Embedded AI collapses that tax—one conversation spans multiple systems, context persists, outputs land where work actually happens.
At Stellis AI, we produce proposals and SOWs constantly. Until recently, each one took 1-2 hours of formatting, hunting, copying, and assembling.
Here's the new workflow:
We keep proposal + SOW templates in a Claude Project
We tell Claude to redline the proposal template with the changes we want to propose to a customer
Claude merges changes into the template, giving a clean Microsoft-Word redline.
We review and tweak it if needed (supervision step)
When approved, we tell Claude to open Gamma and generate a visual proposal using our template
We review and make final adjustments in Gamma (supervision step) and then send to the client
When the client is ready, we tell Claude to convert the proposal into a redlined SOW, using our SOW template.
We review then send for signature
Total time: ~15 minutes unless it's heavily custom.
This isn't "advanced AI implementation." No custom engineering. No platform build. The leverage came from three things: repeatable steps, trusted templates, and system access.
That same pattern applies to contracts, monthly reports, onboarding packs, RFP responses, client updates. If you create similar documents repeatedly, integrated AI turns hours into minutes.
But notice what the workflow actually requires from us: supervision. We instruct. We review. We approve. We catch when AI needs course correction. That's the skill that matters.
Most companies struggle to bridge the gap between "a few people use ChatGPT" and "AI is embedded in our core workflows." Our AI Strategy Action Plan helps leadership teams identify high-impact opportunities, prioritize initiatives, and create a concrete 90-day roadmap that moves from strategy to execution.
The output: a clear plan showing where AI delivers the most value, which workflows to tackle first, and how to build organizational capability systematically—not just chase the latest features.
The AI baseline has shifted. "AI writes emails" is table stakes now. The competitive advantage belongs to people learning to supervise AI working across systems.
Here are three capabilities you need this quarter:
1. Move from prompting to supervising
The skill isn't "write better prompts"—it's "review AI's work across multiple tools and catch what needs correction."
If last week you connected systems (good start), this week focus on building supervision instincts: Which decisions require your review? Which outputs need verification? When should you intervene versus trust the process?
Practice on non-critical workflows. Build guardrails. Document what "good supervision" looks like for your team. This becomes your competitive advantage when agents go autonomous.
2. Expand from 3-4 task types to 7+
Most teams use AI for writing, summarizing, maybe basic research. The gap openers add: data analysis, image work, complex research, translation, specialized tasks unique to their industry.
Pick ONE new task type this month. Not "use AI more"—use AI for something categorically different than what your team does now. That's how you build the capability range that compounds.
3. Build organizational capability, not hero users
Here's the diagnostic: Ask your team, "What recognition have you gotten for using AI well this year?"
If the answer is "none," you have a capability problem disguised as an adoption problem. Research shows 20-45% of employees already use AI regularly, but most feel pressure to hide it.
Stop running "Intro to ChatGPT" training. Start teaching "How to Supervise AI Across Systems," "How to Review AI-Generated Contracts," "When to Trust Automation vs. Dig Deeper." Document supervision standards. Make AI capability visible and valued.
The companies winning aren't waiting for better models. They're building the organizational muscle to leverage autonomous agents before those agents arrive.
Hit reply and tell me: What's the skill gap between where your team is with AI and where they need to be in 18 months? I can tell you what to do today to speed it up.
About Trent: Trent Gillespie is an AI Keynote Speaker, CEO of Stellis AI, former Amazon leader, and advisor on operationalizing AI in business. Book Trent to speak to your group or book a call to discuss using AI within your business.
Tell us what you thought of today's email. |
Did someone forward this newsletter to you? If you're not already signed up, you can subscribe to AI SPRINT™ for free here.