[AI SPRINT] You're Optimizing Prompts. Your Competitors Are Connecting Systems.

Here's what most AI adoption looks like right now:

Your team asks ChatGPT a question. They copy context from Gmail, then Salesforce, then that Google Doc from three weeks ago. They paste everything into the chat. ChatGPT responds. They copy the answer back into Slack.

You call this productivity. It's actually friction with better grammar.

Meanwhile, a small number of companies figured out something different. At Stellis, I ask ChatGPT: "Remind me the background on my 9am meeting tomorrow." It scans my Outlook calendar, pulls the Read.ai summary from our last conversation, checks related HubSpot records, and responds in seconds.

No hunting through email. No checking three systems. One question.

Last month, we built a private ChatGPT app for a client that connects directly to their project management database. Their team uploads an RFP and asks: "Which past projects are most similar, and who on our team has relevant experience?" ChatGPT pulls the answer from their internal systems—no SharePoint archaeology or tribal knowledge required.

That's not a future trend. That's infrastructure that went live in December when OpenAI opened their App Directory to developers.

The Infrastructure Almost Nobody's Using

On December 17th, OpenAI made a huge infrastructure change: they opened app submissions for the ChatGPT App directory, letting any developer publish integrations that run inside ChatGPT conversations. Spotify, Zillow, Uber for consumers. Notion, Slack, Salesforce for business.

But here's what actually matters: ChatGPT can now sit on top of your internal systems and act as the interface to your work.

Not a writing assistant. Not an AI agent. The interface to your other systems.

OpenAI, Anthropic, and Microsoft all launched connector systems in 2025—but I've talked to dozens of companies in the last three months and haven't met ONE using connectors meaningfully. Not one.

The blockers aren't technical:

• Teams are stuck using basic AI features from 2023

• IT doesn't realize how far this has moved

• Leadership isn't aware it exists

• Security concerns freeze anything unfamiliar

• Nobody owns the business case

So companies approve AI, then neuter it. The features that actually change how work gets done never make the roadmap.

Meanwhile, the teams that do connect systems hit a tipping point fast. Once work flows through conversation instead of across six tabs, going back feels absurd.

What Changes When AI Can Actually See Your Work

You've seen the marketing hype: "Let AI summarize today's email!" Sounds useful until you realize you'd still need to read the email to know if the summary matters.

Here's what actually works:

Connect ChatGPT to your calendar, email, and CRM. Ask it to identify customers you haven't contacted in 60 days who might need outreach. It scans all three systems, finds the patterns, drafts the emails.

Connect to Notion. Instead of logging in to check project status, ask: "What's the timeline for the Denver expansion?" ChatGPT pulls context from Notion, cross-references Slack conversations, tells you who's blocked and why.

Connect your meeting notes tool. After every meeting, Read.ai drops summaries into my calendar and HubSpot automatically. When I prep for calls, ChatGPT already has the context. No email archaeology. No "let me find that thread." It's just there.

The win isn't saving 40 minutes a day. It's removing friction from dozens of micro-tasks that used to create drag:

• Finding documents: seconds instead of minutes

• Checking ownership: instant instead of email threads

• Status updates: one question instead of chasing three people

That compounds.

The apps are already live at chatgpt.com/apps (and equivalents in Claude/Copilot).

Connect your systems. Start testing.

One tip: use Thinking mode in ChatGPT. Auto mode moves too fast and misses context.

Five Workflows to Test This Week

These aren't experimental. They work today.

1. Connected project status
Your CEO asks in Slack: "Where are we on urgent items?"
→ "Pull my Asana tasks tagged 'urgent' and draft a leadership update."
(Requires: Asana connector)

2. Meeting prep from real context
You have back-to-back calls and zero prep time.
→ "Check my calendar for this afternoon, pull relevant emails from attendees, and create a meeting brief."
(Requires: Calendar + Gmail/Outlook connectors)

3. CRM-informed outreach
End of quarter. You need to know who to call.
→ "Find Salesforce accounts in tech that haven't heard from me in 60 days and draft personalized check-ins."
(Requires: Salesforce connector)

4. Cross-platform research
Leadership wants product strategy synthesis. It's scattered across tools.
→ "Search Notion and Google Drive for everything on Q4 product launch and summarize our current strategy."
(Requires: Notion + Google Drive connectors)

5. Automated documentation
Engineering wants release notes. You have Slack threads, not documentation.
→ "Pull last week's Slack messages from #product-updates and draft release notes."
(Requires: Slack connector)

The pattern: AI needs to see your actual data, not summaries you manually feed it.

What This Means for 2026

The companies winning with AI won't have the best prompts or the longest policies. They'll have the cleanest connective tissue between AI and their systems.

1. Treat AI like an interface, not a tool

If it can't access live work context, it's decorative. The question isn't "Should we allow ChatGPT?" It's "Which systems should ChatGPT be able to see?"

2. Start with three connectors this quarter

Calendar, email, and one system where work actually lives (Notion, Asana, your CRM). Pilot with a small team. Watch behavior change. The shift from "AI helps sometimes" to "AI is how I work" happens fast once friction disappears.

3. Give your team permission to actually use it 

The biggest blocker isn't technical—it's cultural. Your team needs to know that using AI to pull information from connected systems is the expected workflow, not a workaround. If leadership is still manually checking five systems to answer questions, your team will too.

Model the behavior you want to see.

Analysts predict that by 2028, 60% of knowledge work will happen through conversational interfaces. Not "assisted by" AI—conducted through AI that's connected to the systems where work lives.

Your team will use conversational interfaces for work. The only question is whether those conversations happen inside your systems, or outside them.

Should You Build a ChatGPT App?

You might be wondering now: "Should we build a private app for our team, or a public one for our customers?"

Both are valid. Both require intentional design, security boundaries, and a clear use case.

Some companies should start with private apps—connecting internal data and workflows so teams can actually work through AI instead of alongside it.

Others should build public apps—exposing products, services, or expertise where customers are already spending their time.

The mistake is waiting. Early movers have a 12-18 month advantage before this becomes table stakes.

If you're exploring what a ChatGPT app could look like and want to think it through with someone who's building them in the real world, reach out.

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.

Did someone forward this newsletter to you? If you're not already signed up, you can subscribe to AI SPRINT™ for free here.