How We'll Actually Use AI in Apps
One assistant to rule them all. Not fifty different chatbots.
Jiuhe Wang | Mar 08, 2026
The Mess We're In
Here's how software has worked for decades: you open an app, you do a thing. Excel for spreadsheets. Photoshop for images. Slack for messages. Each app was its own island.
Then AI happened. Now every app has its own chatbot:
I think there's a better way. One assistant. Many capabilities. The apps just do what they're good at.
Two Ways This Could Go
Every app builds its own AI. Different personalities. Different limitations. No shared context.
- Photoshop: "Generative Fill"
- Notion: "Notion AI"
- Excel: "Copilot"
- Slack: "Slack AI"
One assistant orchestrates everything. Apps expose capabilities. Assistant figures out the how.
- One context across all apps
- Apps focus on capabilities, not chat
- Assistant routes requests
- Natural handoffs between tools
Consumer Apps: The Current Chaos
Tuesday morning. You want to draft an email with data from a spreadsheet and a generated image. Here's what happens now:
The Siri Pattern, But Actually Good
Siri already works like this. It doesn't book flights—it opens United. It doesn't order food—it opens DoorDash. The assistant is the brain, apps are the muscles.
Photoshop
Notion
Excel
The assistant:
- Remembers context — knows you were working on the Q2 report
- Routes intelligently> — figures out what "make it prettier" means for each app
- Orchestrates — chains actions without you copy-pasting
Apps Become Capabilities
In the unified model, apps stop being experiences and start being capabilities. Spotify might expose:
You talked. The assistant orchestrated Spotify, found songs, created a playlist, and started your workout tracking. You never opened either app.
Enterprise: Integration Hell
If consumer is messy, enterprise is a disaster. A typical company uses:
Where AI Actually Helps
Traditional integration: mapping fields, writing transformation code, testing, deploying, praying it doesn't break. AI makes it better by understanding what you're trying to do, not just executing commands.
Customer data in Salesforce. Employees in Workday. Money in SAP. None connect.
SAML here, OAuth there, LDAP somewhere else. Plus that spreadsheet of passwords.
REST, SOAP, GraphQL, and custom XML-RPC from a Java app that predates Y2K.
"Revenue" means one thing in Salesforce, another in SAP, another in the CFO's Excel.
AI doesn't make enterprise integration easy—it makes it possible without armies of consultants.
The Enterprise Assistant
An assistant that doesn't just wait for commands—it understands context and proactively gets things done.
✓ Queried Salesforce: $12.4M revenue (+23%)
✓ Pulled Workday: 342 headcount (+28 hires)
✓ Found 3 roadmap docs in Confluence
✓ Created board summary deck
✓ Scheduled exec prep for Thursday 2pm
The assistant saw the meeting, proposed the workflow, executed across all systems, and even flagged anomalies. You just said "yes."
Consumer vs Enterprise
| Factor | Consumer | Enterprise |
|---|---|---|
| Complexity | Dozens of apps | Hundreds of systems |
| Standards | Emerging (Siri, Google) | None (vendor lock-in) |
| Integration | Vendor handles it | Custom per company |
| AI's value | Convenience | Making the impossible possible |
| Time to value | Immediate | Requires setup |
My Guess at the Timeline
Every app has its own AI. ChatGPT, Claude, Copilot—they're islands. You copy-paste between them. Enterprise integration remains expensive and painful.
Siri, Google Assistant deeply integrate with third-party apps. MCP protocol gains rapid adoption. Enterprise AI tools make integrations 10x faster. Early adopters start showing real results.
Apps expose "AI-ready" interfaces as standard. Unified assistant pattern becomes normal. Enterprise platforms go mainstream. Late adopters scramble to catch up.
"Opening an app" feels quaint. You state intent, assistant handles it. App boundaries dissolve. Enterprise fully AI-orchestrated. The old way feels ancient.
My Take
We're in the awkward phase. Every app is building its own AI, creating chatbots that don't talk to each other. It's not sustainable.
The winners will be:
- Apps that are excellent capabilities without pretending to chat
- Assistants that become trusted orchestrators of our digital lives