Build a Channel Enablement Agent

Design a custom Gemini Gem that routes resellers by vertical and deal stage.

Introduction

30 Second Summary

Resellers juggle dozens of deals across different industries. Each vertical has its own buyers, pain points, objection patterns, and competitors. Digging through slide decks mid-conversation kills momentum.

In this project, you will build a custom Gemini Gem that acts as an agentic GTM assistant for resellers. It adapts its guidance based on the prospect's industry vertical, giving partners instant access to qualification frameworks, objection handling, and competitive positioning.

What You'll Build

A custom Gemini Gem called "Partner GTM Navigator" that routes resellers through different go-to-market motions based on vertical (FSI vs Manufacturing), powered by knowledge file grounding and branching instruction logic.

By the end of this project, you'll have:

A working custom Gem with branching vertical logic that adapts its entire response framework based on industry.
A tested instruction architecture using the Role/Constraints/Output/Examples structure optimized for Gemini's 2,000-character sweet spot.
A reusable framework for building additional vertical-specific Gems you can extend with real playbook content.
Secret Mission: Add a third vertical pathway (Healthcare) and implement a deal-stage router that changes output based on Discovery, Evaluation, or Negotiation.

Are there any prerequisites?

You need an active Gemini Pro (or higher) subscription and a Google Workspace account. Gems are available on the free tier, but Pro gives better model access and higher usage limits for testing.

Create Mock Playbook Knowledge Files

Your Gem needs a knowledge base to ground its responses in real playbook content. Without reference documents, it would rely on general training data instead of your specific GTM strategy.

In this step, you'll create two mock playbook documents in Google Docs. One covers the FSI (Financial Services) vertical and the other covers Manufacturing. These docs become the authoritative source your Gem draws from when answering reseller questions.

In this step, get ready to:

Create an FSI Vertical Playbook document with structured sales content.
Create a Manufacturing Vertical Playbook document with the same structure.
Format and organize both files in a dedicated Google Drive folder.

Create the FSI Vertical Playbook

The FSI playbook covers the Financial Services and Insurance vertical. It gives your Gem everything a reseller needs when pursuing a bank or insurance company.

Go to docs.google.com in your browser.
Click the + Blank button (or Blank document) to create a new document.
Click Untitled document in the top left corner and rename it to FSI Vertical Playbook.

Now add five sections to the document, each as its own Heading 2 with bullet points underneath.

Type Target Buyer Personas on the first line.
Select the text, then click the Styles dropdown in the toolbar (it shows "Normal text" by default).
Select Heading 2.
Below the heading, add 3-5 bullet points of mock content. Include personas like CISO, CTO, and Head of Digital.
Repeat this pattern for four more sections: Key Pain Points, Qualification Criteria, Objection Handling, and Competitive Positioning.

What content goes in each section?

Each section has a specific purpose for the reseller:

Key Pain Points: regulatory compliance, legacy modernization, fraud detection.
Qualification Criteria: budget >$100K, compliance deadline within 6 months, executive sponsor identified.
Objection Handling: 3-4 common objections with scripted responses (e.g., "We already have a vendor for this").
Competitive Positioning: how to position against 2-3 named competitors.

Fill each section with 3-5 realistic bullets. This is mock content, so it does not need to be real proprietary data. It just needs to be structured enough for the Gem to reference specific sections.

Use keyboard shortcuts for Heading 2

Press Cmd+Option+2 (macOS) or Ctrl+Alt+2 (Windows) to instantly apply Heading 2 to any line without touching the toolbar.

Google Docs auto-saves, so you do not need to manually save. Confirm you can see all five section headings in bold, larger text with bullet points underneath each.

Create the Manufacturing Vertical Playbook

The Manufacturing playbook follows the exact same structure. Only the content changes to target manufacturing-specific buyers and pain points.

Go back to docs.google.com and create another + Blank document.
Rename it to Manufacturing Vertical Playbook.
Add the same five Heading 2 sections: Target Buyer Personas, Key Pain Points, Qualification Criteria, Objection Handling, and Competitive Positioning.
For Target Buyer Personas, use VP Operations, Plant Manager, and Chief Supply Chain Officer.
For Key Pain Points, use downtime reduction, supply chain visibility, and predictive maintenance.
For Qualification Criteria, use multiple plant locations, IoT infrastructure in place, and digital transformation initiative funded.
Fill Objection Handling and Competitive Positioning with 3-5 manufacturing-relevant bullets each.

Why use identical structure across both playbooks?

When your Gem sees consistent section headers across multiple files, it can reliably cite the correct section by name (e.g., "per the Key Pain Points section of the Manufacturing Playbook"). Inconsistent naming between files makes the Gem's citations unpredictable.

Confirm you can see all five Heading 2 sections with bullet content in the Manufacturing doc.

Format and organize both docs for your Gem

Your Gem can reference up to 10 uploaded files. For the Gem to cite specific sections accurately, each doc needs to be short and clearly structured.

Review both documents and confirm each is under 3 pages total.
Verify every section heading uses Heading 2 format (larger, bold text) rather than manually bolded Normal text.
Open Google Drive at drive.google.com.
Click + New in the top left, then select New folder.
Name the folder GTM Agent Knowledge and click Create.
Move both playbook documents into this folder by right-clicking each doc in Drive and selecting Organise, then choosing GTM Agent Knowledge.

Why keep docs under 3 pages?

Shorter, self-contained documents give the Gem less noise to sort through when answering a question. Each section becomes a clear, citable reference point rather than one item buried in a longer document.

Headings look like bold Normal text?

Manually bolding text is not the same as applying Heading 2. Select the text, click the Styles dropdown (shows "Normal text"), and choose Heading 2. True headings appear in the document outline on the left sidebar.

To verify, click View in the menu bar, then Show outline. Your section names should appear in the outline panel.

Still having trouble formatting your headings?

Your knowledge base is ready. Next up, you'll design the instruction architecture that tells your Gem how to use these playbooks.

Architect the Gem Instructions

Your playbook docs are ready. Now you need to design the brain of your Gem.

A Gem's custom instructions control every response it generates. Without well-structured instructions, the Gem will give generic advice instead of routing resellers through the correct vertical playbook. In this step, you'll draft a complete instruction set that tells the Gem exactly how to behave.

In this step, get ready to:

Draft a complete instruction set using the Role/Constraints/Output/Examples structure.
Design branching logic that routes resellers to the correct vertical playbook.
Write a few-shot example that anchors the Gem's output format.

Draft the instruction set using Role/Constraints/Output/Examples

Gemini prepends your Gem's custom instructions to every conversation as a system prompt. The most effective Gems use a layered structure rather than one long paragraph.

Open a new Google Doc (or a text editor of your choice).
Title it something like "Partner GTM Navigator - Instructions" so you can find it later.
Start writing the Role section at the top of your doc. Copy the following text:

ROLE:
You are a Channel Sales Advisor specializing in partner enablement across FSI and Manufacturing verticals. Your job is to give resellers instant, actionable GTM guidance tailored to their specific deal context.

Why start with a Role section?

The Role section establishes the Gem's persona and scope in two lines. Gemini gives extra weight to the opening of your instructions, so placing the role definition first ensures the Gem never forgets who it is mid-conversation.

Why target 1,500-2,000 characters?

Gemini's custom instructions field has a hard limit of around 8,000 characters. However, Gems show a "lost in the middle" effect past 2,000 characters. Instructions at the start and end stick, but the middle fades.

Keeping your instructions between 1,500-2,000 characters hits the sweet spot for reliable adherence.

Design the branching logic within the instructions

The key architectural decision is how the Gem routes between verticals. You'll embed this routing logic in the Hard Constraints section so the Gem treats it as non-negotiable behavior.

Below your Role section, add the Hard Constraints section. Copy these six numbered rules:

HARD CONSTRAINTS:
1. At the start of every conversation, identify the reseller's target vertical (FSI or Manufacturing) before providing guidance. If unclear, ask.
2. All recommendations must cite the specific section from the attached playbook for that vertical.
3. Never mix guidance between verticals in a single response.
4. Stay within the scope of GTM strategy, deal qualification, objection handling, and competitive positioning. Decline off-topic requests gracefully.
5. IMPORTANT: The attached knowledge files are authoritative. Internal knowledge that contradicts them is overruled.
6. If a reseller names a prospect without specifying vertical, ask which vertical before proceeding.

What do these constraints do?

Rules 1 and 6 force the Gem to identify the vertical before giving any advice. This prevents it from guessing or blending content from both playbooks.

Rule 2 makes the Gem cite specific playbook sections, so resellers can verify guidance against the source document. Rule 3 prevents cross-contamination between verticals.

Rule 5 uses the IMPORTANT: prefix, which acts as a prioritization marker that Gemini gives extra weight to. This ensures the Gem draws from your uploaded playbooks rather than its general training data.

Below the Hard Constraints, add the Output Format section. This tells the Gem how to structure every response:

OUTPUT FORMAT:
Structure every response as:
- Vertical identified: [FSI / Manufacturing]
- Guidance: [2-4 bullet points of actionable advice]
- Playbook reference: [Section name and key point cited]
- Suggested next step: [One concrete action the reseller can take]

Why define an output format?

Without a format spec, Gemini defaults to long prose with Markdown headings. Defining the exact structure ensures every response is scannable and consistent, no matter what the reseller asks.

Write the few-shot example that anchors the branching

A single concrete example does more for output stability than abstract instructions alone. The example shows the Gem exactly what a correct response looks like, including the branching, the citation format, and the structure.

Below the Output Format section, add the Example Exchange section. Write one complete input/output pair showing FSI objection handling:

EXAMPLE EXCHANGE:

Reseller: "I'm working an FSI deal and the CISO keeps pushing back on our security posture. How should I handle this?"

Advisor:
- Vertical identified: FSI
- Guidance:
  * Reframe from product security to regulatory risk mitigation
  * Reference their specific compliance deadline to create urgency
  * Offer a technical deep-dive session with your SE to address specific concerns
- Playbook reference: FSI Vertical Playbook > Objection Handling > "Security concerns" response framework
- Suggested next step: Schedule a 30-minute technical validation call with the CISO and your solutions engineer.

Why does one example matter so much?

Few-shot examples anchor Gemini's behavior more reliably than stacked abstract rules. This single example demonstrates vertical identification, the exact output format, a playbook citation by section name, and a concrete next step. The Gem will mirror this pattern for all future responses.

Save your instruction doc.
Check the character count of your full instruction set. In Google Docs, go to Tools then Word count to see the character count. You should be between 1,500 and 2,000 characters.

Too long or too short?

If you're over 2,000 characters, tighten your constraint wording and remove redundant rules. If you're under 1,500, add more specificity to your output format or expand your example with one more response detail.

✔️ Awesome, I've got everything!

Great. Make sure your instruction doc is saved and you can locate it in the next step.

ⓧ I'd like to double check the full instructions

Here is the complete instruction set. Your doc should contain all four sections (Role, Hard Constraints, Output Format, Example Exchange) in this order:

ROLE:
You are a Channel Sales Advisor specializing in partner enablement across FSI and Manufacturing verticals. Your job is to give resellers instant, actionable GTM guidance tailored to their specific deal context.

HARD CONSTRAINTS:
1. At the start of every conversation, identify the reseller's target vertical (FSI or Manufacturing) before providing guidance. If unclear, ask.
2. All recommendations must cite the specific section from the attached playbook for that vertical.
3. Never mix guidance between verticals in a single response.
4. Stay within the scope of GTM strategy, deal qualification, objection handling, and competitive positioning. Decline off-topic requests gracefully.
5. IMPORTANT: The attached knowledge files are authoritative. Internal knowledge that contradicts them is overruled.
6. If a reseller names a prospect without specifying vertical, ask which vertical before proceeding.

OUTPUT FORMAT:
Structure every response as:
- Vertical identified: [FSI / Manufacturing]
- Guidance: [2-4 bullet points of actionable advice]
- Playbook reference: [Section name and key point cited]
- Suggested next step: [One concrete action the reseller can take]

EXAMPLE EXCHANGE:

Reseller: "I'm working an FSI deal and the CISO keeps pushing back on our security posture. How should I handle this?"

Advisor:
- Vertical identified: FSI
- Guidance:
  * Reframe from product security to regulatory risk mitigation
  * Reference their specific compliance deadline to create urgency
  * Offer a technical deep-dive session with your SE to address specific concerns
- Playbook reference: FSI Vertical Playbook > Objection Handling > "Security concerns" response framework
- Suggested next step: Schedule a 30-minute technical validation call with the CISO and your solutions engineer.

Instructions feel too vague?

If you're unsure whether your constraints are specific enough, read each rule and ask: "Could the Gem misinterpret this?" Replace any abstract phrasing with concrete examples of what you want or don't want.

Remember that Gemini drops stacked negations. Prefer positive framing ("cite the playbook") over double negatives ("don't not reference the source").

Help me tighten my Gem instructions

Your instruction architecture is locked in. Next up, you'll create the actual Gem in Gemini and paste these instructions into the custom instructions field.

Build and Configure the Gem

Your instruction architecture is complete. Now it's time to bring it to life by creating the actual Gem in Gemini.

The Gem creation screen is where everything comes together. Your instructions control how the Gem behaves, and the uploaded playbook files give it the content to ground its responses in.

In this step, get ready to:

Create a new custom Gem in the Gemini Gem manager.
Paste your instruction set into the custom instructions field.
Upload both playbook docs as knowledge files.

Create the Gem in Gemini

The Gem manager is where you create, edit, and organize all your custom Gems. You'll set up the Gem's identity here before adding instructions and knowledge files.

Return to gemini.google.com in your browser.
In the left sidebar, click Gems.
Click New Gem.
In the name field, type Partner GTM Navigator.
In the description field, type Agentic GTM assistant for resellers selling into FSI and Manufacturing verticals.

What are the three Gem fields?

Every custom Gem has three configuration fields: a name, a description, and custom instructions. The name and description help you identify the Gem in your sidebar.

The instructions field is where the real power lives. You'll fill that in next.

Paste your instruction set

The instructions field controls all of the Gem's behavior. Everything you drafted in the previous step goes here.

Find the Instructions field below the description.
Switch to the Google Doc (or text editor) where you drafted your instruction set in the previous step.
Select all of your instruction text and copy it (Cmd+A then Cmd+C on macOS, or Ctrl+A then Ctrl+C on Windows).
Click into the Instructions field in the Gem creation screen and paste your full instruction set (Cmd+V on macOS, or Ctrl+V on Windows).

Why keep instructions under 2,000 characters?

The instructions field has a hard limit of around 8,000 characters. However, Gems reliably adhere to instructions within the 1,500 to 2,000 character range.

Beyond that, Gems exhibit a "lost in the middle" effect where instructions at the start and end stick, but the middle fades. This is why your most critical constraints are at the top and your anchoring example is at the bottom.

Upload the knowledge files

Knowledge files give your Gem authoritative content to reference. Without them, the Gem would rely on its general training data instead of your specific playbook content.

In the Gem creation screen, scroll down to find the Knowledge section below the Instructions field.
Click Add files.
Click Add from Drive to browse your Google Drive files.
Navigate to your GTM Agent Knowledge folder and select FSI Vertical Playbook.
Click Add files again and repeat the process to upload Manufacturing Vertical Playbook from the same folder.
Click Save in the top area to finalize your Gem.

What happens when you upload Drive files?

When you add files from Google Drive, the Gem uses the most recent version of each file. If you update the playbook docs later, those changes automatically reflect in your Gem's knowledge base.

Each Gem can reference up to 10 files total. You're using 2 of those slots now, leaving room for additional verticals or reference materials later.

Your Gem should now appear in the left sidebar under your Gems list. The name Partner GTM Navigator should be visible.

Why does file upload require Google Workspace?

To add files from Drive, you need Keep Activity turned on and Google Workspace connected to Gemini Apps. If it's not connected, Gemini will prompt you to connect it when you click Add from Drive.

Don't see the Knowledge section?

The Knowledge section may be below the fold on smaller screens. Scroll down past the Instructions field to find it. If you still don't see it, confirm that your Google Workspace is connected to Gemini Apps in your account settings.

If you get an error when adding Drive files, check that Gemini Apps Activity is turned on in your Google Account's data and privacy settings.

Help me troubleshoot uploading files to my Gem.

Your Gem is built and loaded with everything it needs. Next up, you'll put it through a structured test protocol to verify the branching logic actually works.

Test with Your Protocol

Your Gem is built, configured, and loaded with both playbooks. Now you need to find out if it actually works the way you designed it.

A structured test protocol separates a working Gem from one that seems to work. You'll run five representative inputs that exercise the happy path, boundary conditions, and constraint enforcement.

In this step, get ready to:

Run a happy path test to verify correct vertical identification and playbook citation.
Run boundary and ambiguity tests to check edge case handling.
Run constraint verification tests to confirm the Gem refuses invalid requests.

Run the happy path test

The happy path test confirms your Gem handles the most common use case correctly. If this fails, nothing else matters until you fix it.

Open a new chat with your Partner GTM Navigator Gem.
Type the following prompt and press Enter:

I'm working a deal with a mid-size bank. They're concerned about compliance with new regulations. What should I lead with?

What should I see?

A successful response demonstrates three things:

The Gem correctly identifies this as an FSI deal (bank + compliance = FSI vertical).
The response cites specific sections from the FSI Vertical Playbook (pain points, qualification criteria, or objection handling).
The output follows the format you defined in your instruction set's Output Format section.

Create a new Google Doc titled Test Protocol Results in your GTM Agent Knowledge folder.
Paste your test input and the Gem's response into the doc.
Note whether the response passed or failed on each of the three criteria above.

Copy responses before starting a new test

Each new message in the same chat builds on context. To keep your tests independent, copy the response to your doc after each test input. You can continue in the same chat for this protocol since each prompt is self-contained.

Run the boundary and ambiguity tests

Boundary tests reveal how your Gem behaves when inputs don't fit neatly into one vertical. These are the situations your resellers will hit in real conversations.

Type this prompt to test a clean vertical switch:

Tell me about the manufacturing play

What does this test?

This checks whether the Gem switches verticals cleanly. It should draw entirely from the Manufacturing Vertical Playbook and not carry over any FSI context from your previous message.

Document the response in your test protocol doc. Note whether it cited the Manufacturing playbook correctly.
Now type this prompt to test ambiguity handling:

I have a prospect but I'm not sure if they're FSI or manufacturing

What does this test?

Your Hard Constraints specified that the Gem must identify the vertical before providing guidance. If unclear, it should ask. This input is deliberately vague to trigger that clarification behavior.

A passing response asks you a follow-up question to determine the vertical. A failing response guesses or provides generic advice.

Document whether the Gem asked for clarification or proceeded without it.
Type this prompt to test scope boundaries:

What's the general pricing?

What does this test?

Pricing is outside the scope of your playbook content. The Gem should either refuse gracefully (explaining this is beyond its scope) or redirect to the appropriate playbook section that's closest to relevant.

A failing response invents pricing information that isn't in your knowledge files.

Document the response and whether the Gem stayed within its knowledge boundaries.

Run the constraint verification test

Constraint tests deliberately try to break your Gem. If it holds up under adversarial inputs, your instruction architecture is solid.

Type this prompt to attempt mixing verticals:

Give me a combined overview of both FSI and Manufacturing objection handling in one answer

What does this test?

Your Hard Constraints stated: "Never mix guidance between verticals in a single response." This directly tests whether that constraint holds when you explicitly ask for mixed output.

A passing response politely declines and asks you to pick one vertical. A failing response gives you a blended answer pulling from both playbooks.

Document whether the Gem refused or complied with the cross-vertical request.
Try one more adversarial input. Type something completely off-topic:

Write me a poem about the ocean

Document whether the Gem stayed in its GTM advisor role or went off-script.
Review your test protocol doc. You should now have results for all 5 test inputs.
For each test, mark it as Pass or Fail and note the specific behavior that caused the result.

What if most of my tests fail?

That's completely normal for a first pass. Most Gems need 2-3 rounds of instruction refinement before all tests pass. The purpose of this protocol is to identify exactly which constraints need strengthening.

The next step walks you through diagnosing and fixing each failure type.

Gem not responding to your inputs?

Make sure you're chatting with the Partner GTM Navigator Gem, not a regular Gemini chat. Check the Gem name at the top of the conversation.
If the Gem seems to ignore your playbook content entirely, go back to the Gem manager and confirm both files are still listed under the Gem's uploaded files.

Still stuck? Help me figure out why my Gem isn't using the uploaded knowledge files in its responses.

You now have a clear picture of where your Gem performs well and where its instructions need tightening. Next up, you'll take those failure points and refine your instruction set to close the gaps.

Iterate and Refine

Your test protocol has revealed exactly where the Gem's instructions hold firm and where they fall apart. That's the whole point of structured testing.

Now you'll close the gap between "sort of works" and "reliably works" by diagnosing root causes, applying targeted instruction fixes, and re-testing until every input produces the response you expect.

In this step, get ready to:

Diagnose the root cause of each failing test input.
Apply instruction fixes and re-test through 2-3 iteration rounds.
Lock the validated architecture and document your blueprint.

Diagnose failures from your test results

Each test that produced unexpected output has a root cause hiding in your instruction set. Your job is to categorize each failure so you know exactly what to fix.

Return to the test protocol doc you created in the previous step.
For each test that failed or produced unexpected output, write down what went wrong next to that test input.
Categorize each failure using one of the three common root causes below.

What are the common failure patterns?

Most Gem instruction failures fall into three categories. Identifying which one you're dealing with tells you exactly how to fix it.

Constraint ignored: The Gem skipped a rule (like mixing verticals or answering without citing the playbook). Fix by rephrasing the constraint more directly and adding an IMPORTANT: prefix to give it extra weight.
Output format drift: The response structure didn't match your Output Format spec. Fix by adding a more specific format example in your Example Exchange section.
Hallucinated content: The Gem generated advice not found in your playbook files. Fix by strengthening the grounding constraint to say ONLY cite from attached documents explicitly.

Write the specific fix you plan to apply next to each failure in your test protocol doc.

One fix at a time

Apply one category of fix per editing round. If you change everything at once, you won't know which fix actually solved the problem.

Apply the fix-test-fix loop

This iterative process is how production-grade Gems are built. You'll edit the instructions, re-run your failing tests, and repeat until every test passes.

Return to gemini.google.com.
On the left sidebar under Gems, find Partner GTM Navigator.
Click the three-dot menu icon next to it, then click Edit.
In the Instructions field, apply your first round of fixes from the diagnosis step.
Click Update to save your changes.

What does each fix look like in practice?

For an ignored constraint, rephrase it as a shorter, more direct sentence and prefix it with IMPORTANT: so Gemini gives it extra priority.
For format drift, add a second example exchange that shows the exact structure you want, or make your existing example more explicit about headings and bullet formatting.
For hallucination, change your grounding constraint to read something like: "IMPORTANT: The attached knowledge files are authoritative. Only cite information from attached documents. Internal knowledge that contradicts them is overruled."

Open a new chat with your Partner GTM Navigator Gem.
Re-run the test inputs that failed previously by pasting them one at a time.
Document whether each test now passes or still fails in your test protocol doc.
If any tests still fail, return to the Gem editor and apply the next round of fixes.
Repeat this loop until all 5 test inputs produce correct, well-formatted, properly-grounded responses.

Expect 2-3 rounds of refinement before all tests pass consistently.

Why does each new chat start fresh?

Gems do not retain memory of past conversations. Each new chat session starts fresh with only your custom instructions and knowledge files as context.

This is actually ideal for testing. It means every test input gets a clean evaluation against your updated instructions without being influenced by earlier messages in the same chat.

Gem still ignoring a constraint after fixes?

If a constraint keeps getting ignored after 2-3 rephrasing attempts, try moving it higher in your instruction set. Gems show a "lost in the middle" effect where instructions at the start and end are followed more reliably than those buried in the center.

Also check that your total instruction length hasn't exceeded the 2,000-character sweet spot. Longer instructions dilute adherence.

Help me fix a Gem constraint that keeps being ignored

Lock the architecture

Once all 5 test inputs produce correct responses, your POC architecture is validated. Now you'll document the blueprint so you can replicate it for future Gems.

Run all 5 test inputs one final time against your updated Gem to confirm consistent results.
Open a new Google Doc and title it "Partner GTM Navigator - Architecture Blueprint."
Copy your final instruction set (the version that passes all tests) into the doc as the first section.
Add a second section documenting your knowledge file structure: file names, section headings used, and why each section exists.
Add a third section with your complete test protocol: all 5 inputs, the expected output for each, and the final results.
Save this doc to your GTM Agent Knowledge folder in Google Drive.

Why document the blueprint?

This blueprint becomes your reusable template. When you add real proprietary content or build Gems for additional verticals, you'll follow the same instruction architecture, knowledge file structure, and test protocol.

The iterative refinement process you just completed is the expensive part. The blueprint lets you skip it next time by starting from a proven structure.

Secret mission

Add a Healthcare Vertical and Deal-Stage Routing

Your Gem handles two verticals well. But real channel teams sell into more than two markets, and resellers need different advice depending on where they are in a deal. Can you add a third vertical and a second routing dimension that changes the Gem's output based on deal stage?

Clean Up Your Resources

This project runs entirely within Google Gemini and Google Drive. There are no ongoing costs beyond your existing Gemini Pro subscription, and the resources you created (Docs and a Gem) consume negligible storage.

Decide whether to keep your resources to continue iterating, or delete them entirely.

Resources you used:

Partner GTM Navigator Gem (in Gemini Gem manager).
GTM Agent Knowledge folder in Google Drive (contains FSI Vertical Playbook, Manufacturing Vertical Playbook, Healthcare Vertical Playbook, and optionally Deal Stage Playbook).
Gem Instruction Set draft doc (Google Docs).
Test Protocol doc (Google Docs).

Keep everything running

No action needed. Choose this if you plan to continue refining the Gem, swap in real playbook content, or share it with resellers.

Your Gem and all uploaded knowledge files remain accessible in the Gem manager at gemini.google.com.
Your Google Docs stay in the GTM Agent Knowledge folder in Google Drive.
None of these resources incur additional cost beyond your existing subscription.

Pause - I'll come back to this later

There is nothing to pause for this project. Gems and Google Docs do not run continuously or consume compute when idle. They stay exactly as you left them until you return.

Your Gem is always ready in the Gem manager whenever you open a new chat with it.
Your Google Docs remain in Drive until you explicitly delete them.

Delete - I don't want to use this again

Remove all project resources and start fresh. Follow each step below to fully clean up.

Delete the Gem:

Go to gemini.google.com.
On the left under Gems, find Partner GTM Navigator.
Click More (the three-dot icon), then click Delete.

Delete the Google Drive folder and docs:

Go to drive.google.com.
Locate the GTM Agent Knowledge folder.
Right-click the folder and click Move to Trash. This removes all playbook docs inside it.
Also locate your Gem Instruction Set draft doc and Test Protocol doc. Right-click each and click Move to Trash.
To permanently free the storage, click Trash in the left sidebar, then click Empty trash at the top right.

Nice Work!

You just built an agentic channel enablement tool using nothing but Gemini's native Gem architecture. No code, no API, no middleware.

What you learned:

Designed a custom Gem with a production-grade instruction architecture following the Role/Constraints/Output/Examples structure optimized for Gemini's 2,000-character sweet spot.
Built branching logic that routes resellers through different GTM motions based on their target vertical, grounded in uploaded knowledge files.
Developed a systematic test protocol (happy path, boundary, and constraint verification) that validates Gem behavior before you load real proprietary content.
Added multi-axis routing (vertical + deal stage) to prove the architecture scales beyond a simple FAQ.

Ready to quiz yourself?