How to Delegate Work to Human Experts from Claude or ChatGPT

You've felt this already if you use Claude or ChatGPT daily. Some tasks land. Others get you 70% of the way and stall. The research that sounds confident but you're not sure you can use. The contact list with three names that don't actually work at those companies. The pitch deck feedback that could apply to anyone's pitch deck. The copy that you keep rewriting because something's off.

Until recently, the only fix was to give up on the AI chat for that part of the work, find a freelancer, brief them from scratch, and copy results back. With the Tendem MCP connector, a different option exists: you delegate the task to a vetted human expert without ever leaving the chat. The AI handles scoping and routing; a real person delivers the result; everything lands back in the same conversation.

This guide walks through how to actually do it well – setup, the prompt patterns that get the best results, what the back-and-forth looks like, and the simple decision rules for when to delegate versus when to keep going on your own.

Step 1: Connect Tendem to your AI tool

Setup takes about a minute. The mechanics are the same in every supported tool (Claude, ChatGPT, Claude Cowork, Cursor, Codex, and most other MCP-enabled AI tools):

1. Open the connector settings. In Claude: Settings → Connectors → Add Custom Connector.

2. Name it Tendem, paste the server URL https://mcp.tendem.ai/mcp

3. Click Add → Connect, authorize via OAuth, and you're set.

Full platform-specific instructions live on the Tendem MCP setup page. Once connected, you simply mention Tendem in your prompt and the agent picks up the tool automatically.

Step 2: Write a brief the AI can scope

The single biggest determinant of how well delegation works is how concrete your initial task is. Vague briefs ("help me with marketing") produce vague scoping back-and-forth and unpredictable pricing. Specific briefs with a clear deliverable get clean prices and clean deliverables.

A good Tendem prompt has four parts:

• The verb. What action do you want done? Research, find, build, write, review, clean, enrich, verify.

• The subject. What's the topic, list, document, or asset?

• The constraints. Quantity, geography, industry, criteria, timeline, accuracy bar.

• The deliverable format. Spreadsheet, Google Doc, report, slide review with comments.

Compare the two:

The agent will ask clarifying questions on vague briefs – it's trying to tighten the brief before submitting. Saving that round of back-and-forth is what front-loading specificity buys you.

Step 3: Prompt patterns by task type

Five patterns cover most real-world Tendem tasks. Use them as templates.

Research and competitive analysis

"Use Tendem to research [topic/competitor set]. I want [3-5 specific data points] for each. Deliverable: [format]. Cite sources where applicable."

Works well for: competitive teardowns, market sizing, vendor evaluations, regulatory landscapes, anything where you'd otherwise spend a day reading and synthesizing.

List building and data enrichment

"Use Tendem to find [N entries] matching [specific criteria]. For each, include [list of fields]. Geography: [region]. Deliverable: spreadsheet."

Works well for: prospect lists, investor lists, partner shortlists, conference attendee follow-ups. The accuracy gap between this and AI-generated lists is the difference between a usable file and a dead-end exercise – we wrote about that gap in human data verification.

Content and copy

"Use Tendem to write [N pieces] of [content type] based on this [source material]. Tone: [tone]. Length: [length]. Audience: [audience]. Include [specific asks like CTAs, hooks, hashtags]."

Works well for: outreach sequences, LinkedIn posts, blog drafts, proposals, anything where an AI draft would land at 70% but needs the last 30% from a human who actually understands the audience.

Design and creative review

"Use Tendem to review [asset – attach or link]. I want [type of review – e.g. design hierarchy, brand fit, copy clarity]. Call out what doesn't work and suggest fixes."

Works well for: pitch decks before they go out, landing page mockups, brand asset feedback, presentation polish. You can ask for two perspectives in the same brief and Tendem will route to two specialists in parallel.

Multi-expert / complex tasks

"Use Tendem to review this pitch deck. I want a designer's eyes on the layout and visual hierarchy, and a copywriter checking the messaging is on point. Call out what doesn't work and suggest fixes."

The agent reads this as two distinct tasks, opens both in parallel, and brings the results back into the same conversation when each is ready. You don't have to organize anything.

Step 4: The scoping and pricing round

After you submit, three things happen, all in chat:

1. Clarifying questions, if needed. If the brief has gaps, the agent will ask one or two targeted follow-ups before sending it for pricing. Answer them in chat and the brief gets resubmitted automatically.

2. Pricing. Tendem reviews and prices the task within about 10 minutes. The proposal appears in chat as a price plus a short summary of what's included.

3. Your call. Approve the price and an expert gets started. If the price feels high, ask the agent to narrow the scope ("just the top 5 competitors, not 10") and resubmit. If you walk away, nothing is charged – nothing leaves your account until you say yes.

This is the trust mechanism that makes the whole thing work for first-time users. You see the number before you commit, and you can negotiate the scope down without re-briefing from scratch.

Step 5: Track progress and receive the result

After you approve, the expert works. You can keep using your AI tool for everything else – the task runs in the background. To check status, just ask: "check my Tendem task and tell me if it needs my approval."

When the work is done, results return to the same conversation as readable text plus downloadable files, with a link to the full task in Tendem. You can immediately ask the AI to summarize, restructure, or work on top of the deliverable – it has the full output in context.

When to delegate vs. when to keep going on your own

Not everything needs a human. The point of having the connector available is precisely so you can make the call task by task. A simple rule of thumb:

Our deeper guide on when to use human experts instead of AI walks through the decision in more detail.

Common mistakes (and how to avoid them)

Submitting a vague brief and accepting whatever the agent scopes. If the agent's clarifying questions feel like a guessing game, stop and rewrite the brief with the four parts above. You'll get a faster price and a result closer to what you actually need.

Treating the price as fixed. Pricing is based on the scope you submitted. If it feels high, the answer is almost always to narrow the brief, not to look elsewhere. "Reduce scope to the most important 5 of those 15 entries" is a perfectly normal next prompt.

Forgetting deliverable format. Specifying "spreadsheet with these columns" or "Google Doc with these sections" prevents 80% of the friction at handoff. The expert delivers exactly what's usable, not a free-form write-up you'd have to reformat.

Trying to delegate everything. Tendem is a power tool for the work that genuinely benefits from human judgment. Human-in-the-loop systems work best when you're clear about which loop they're closing – not every task is a fit.

Try it on a real task today

The fastest way to understand the workflow is to run a small task through it. Pick one thing on your list this week that your AI tool can't quite finish on its own – a research synthesis, a 20-row list build, a copy review – and hand it to Tendem in the chat where you'd normally start it.

First-time users get a $50 starting bonus, and the first three tasks are 50% off. Connect Tendem to your AI tool to get started.

Related Resources

• Human-in-the-Loop: The Complete Guide – the architecture behind AI + human delegation

• When to Use Human Experts Instead of AI – decision rules in more depth

• Human Data Verification: AI Output, Expert Review – why verification changes deliverable quality

• The True Cost of AI Hallucinations in Business Data – what AI alone gets wrong, and what it costs

• Outsource Repetitive Tasks Without Managing Freelancers – the freelancer-overhead problem Tendem solves