Tips for getting started with Gemini Gems: How to Build Custom AI Agents
If you can get a subscription for fancy LLM tools and they will solve most of your operational and marketing woes, you are signing up for a big blunder. The magic of AI integration is not in using LLMs directly, but it’s in tools tailored to your workflows.
As marketing leaders, we’ve all used content generators for producing various forms of proprietary content, but they don’t align with your brand tone, even if you instruct them with guidelines. They’re entertaining but not actionable.
Marketing and sales teams still spend hours manually replying to emails, summarising reports and hunting for information across multiple systems.
This leads us to ask this common question: Why can’t we have an AI assistant that knows our tone, pulls live data from our tools and follows our rules?
That’s exactly what Google’s new Gemini Gems offers.
Gems are custom AI agents that live inside the Gemini platform and follow your instructions. It can act as your content expert, sales manager and a full-stack coding partner, all of these at the same time.
In this post, we’ll share how to build these agents, explore real use-cases in BFSI, healthcare, technology and finance, and lay out the opportunities and risks for enterprises.
Step-by-step: Creating a Gem
- You need to upgrade to Gemini Advanced. Without it, you can’t use Gems.
- Now, open Gem Manager (left sidebar → Explore Gems) and click New Gem.
- In one sentence, describe what it should do. For example: “Reply to emails in a friendly yet professional tone.”
- Click on Prompt Generator. Gemini expands your one-liner into a detailed instruction set with purpose, tone, goals and constraints.
- Turn on relevant extensions as per your task preference (Gmail, Drive, Docs, YouTube). In the instruction or chat, you can use @Gmail, @Drive or @Docs to pull specific emails or documents into the conversation.
- Try a few sample queries in the preview pane, adjust the prompt if needed, then click Save.
- Your Gem is now ready and will follow its instructions without you having to repeat the prompt.
By following these steps, you go from a broad idea to a functioning agent that understands your tone and uses live data. Next, let’s examine why this matters for key industries.
This is how a Gemini gem window looks:

Exploring Gems in the Gem Manager

Google provides a set of ready-made Gems that show practical use cases. You can use it before designing your own AI agents around proprietary workflows:
- Storybook - Create customised picture books for children or adults, based on your chosen topic and style.
- Chess Champ - Play chess with a language model that understands moves and strategies.
- Brainstormer - Generate fresh ideas quickly, from business concepts to party themes.
- Career Guide - Get personalised career advice and development plans.
BFSI: Risk, Compliance and Personalised Customer Support
While machine learning already supports risk modelling and chatbots, those bots still need frequent prompting and don’t understand firm-specific policies.
A BFSI Risk & Compliance Gem can solve this by connecting directly to internal risk reports. It can summarise filings, highlight compliance red flags, flag anomalous transactions, and generate a management briefing in the right tone, all without repeated instructions.
Why this matters
- Risk and compliance efficiency - By ingesting policies and regulatory updates, a Gem can automatically identify high-risk transactions, summarise regulator notices and generate tailored compliance emails.
- Better customer advice - A Gem could use your firm’s research database to answer client questions in seconds.
- Fraud detection - Generative models can produce mock fraud examples to train detection systems and improve compromised card detection.
We’re still early, yet the stakes are high. Leaders must ensure Gems handle sensitive financial data securely and explain their decisions clearly to comply with audit requirements.
Healthcare: clinical summaries and patient engagement
Healthcare workers battle with paperwork and fragmented patient data. Over 79 % of providers already use AI, seeing ROI within 14 months. Gems can bring the power of generative AI directly into clinicians’ workflows.
A Healthcare Clinical Assistant Gem can do all of these -
- Connects to electronic health records
- Summarises patient history
- Flags potential drug interactions
- Drafts a physician's briefing
- Maintains neutral tone
- Highlights actionable insights
How healthcare teams benefit
- Clinical decision support - AI summarises lengthy patient records while highlighting abnormal lab results, and also suggests relevant clinical trials.
- Compliance and privacy - Only authorised users can access the Gem, and data never leaves the organisation.
- Patient engagement - A Gem trained on your FAQs and discharge instructions can answer patient questions in plain language.
Tech Field: Powering software delivery and DevOps
Developers waste hours triaging bug reports and searching across documentation. Gems can help teams move faster while keeping humans in control.
The DevOps Assistant Gem shown above pulled bug reports from the Jira board, highlighted critical issues and assigned them to the right teams.
What this means for tech leaders
- Accelerated debugging and review - It proposes fixes, refactors code, identifies vulnerabilities, and triages tickets.
- Stronger DevOps pipelines - It analyses build logs, predicts failures, and recommends optimisations to speed releases.
- Upskilling, not replacement - It empowers programmers to work as a co-pilot rather than replacing them.
Finance And Corporate Reporting: From Insights To Action
Beyond banking, finance teams across industries manage budgets, forecasts and investor communications.
Portfolio Insights Gem can analyse a client’s holdings, identify over-exposure to tech stocks, recommend rebalancing, and generate a concise report saved directly to the advisor’s Drive.
With a Gem, financial advisors don’t need to craft explanations from scratch; the tone and analysis stay consistent with your brand.
Applications
- Portfolio analysis - Summarises holdings, flags risk concentration, and suggests diversification.
- Earnings calls - Drafts scripts, anticipates questions, and generates responses from past transcripts.
- Scenario simulation - Models macroeconomic shifts, forecasts asset prices, and stress-tests with synthetic data.
Finance leaders must maintain oversight and ensure the Gem complies with regulatory frameworks.
Challenges
- Data privacy - Live access to Gmail and Drive requires strict access controls and audit logs.
- Reliability - Gems can hallucinate; human review remains critical.
- Cost & lock-in - Gemini Advanced is paid, and relying on Google’s ecosystem may limit flexibility.
Turning Gemini Gems into Competitive Edge
Custom AI agents will become the nerve centre of enterprise marketing and operations. Unlike generic chatbots, a Gem trained on your proprietary data and tone becomes a trusted colleague.
Imagine connecting your CRM, campaign reports and customer feedback to a Gem that surfaces actionable insights every morning.
Are you ready to design your own Gem?
Do you want more traffic?

Your Brand, Your Rules. Leading in the Age of AI

Your Brand is Failing Without This: Why Omnichannel is the ONLY Strategy That Matters

STOP Targeting Google: The "Search Experience" Trick That's Quietly Doubling Traffic
