#86: AI Agents Unlock What Your Data’s Been Hiding
The best part of our AI Accelerator is seeing the lightbulb moment when clients see just how powerful AI becomes once it’s connected to their own data. Sure, we’ve all tossed a slide deck or an Excel sheet into a prompt and asked it to read it or summarize, but have you asked ChatGPT to provide a list of your biggest clients, the products they’ve purchased, and the salesperson for each of them? Have you asked a model to look at the activity by salesperson across your CRM and help build suggestions to increase the sales performance for each of them?
Thankfully, you can start to get a taste of this in the connectors that Anthropic and OpenAI started rolling out for paid tiers. You can find these by going to your settings, selecting the connectors option, and it will list applications that have created MCP bridges (read up on MCP and RAG if you missed those), allowing an AI agent to have a conversation or access these data sources as knowledge bases.
In the last few months, HubSpot rolled out experimental access to an MCP server that gives deep querying access to your HubSpot data, and I find it better than trying to use its reporting features to query and get slices of data. You can also build your own MCP servers, something my team does for many organizations to unlock the ability for your CFO to have a conversation with Sage or NetSuite with natural language, or for your product teams to talk directly with inventory and supplier data in your ERP.
I’ve started to produce deals and sales reports using AI tools connected with my business data. Reports that you would have to dig through dashboards and export files have turned into simple conversations. Instead of clicking through multiple screens to understand pipeline health or rep performance, I can now ask Claude directly.
So to demo this, I integrated Claude with HubSpot.
Step 1: In Claude, go to Settings and click on “Connectors”. You’ll see a library of applications that have built MCP integrations, including HubSpot, Google Drive, Gmail, Asana, and dozens of others.
Step 2: Toggle on HubSpot (and any other tools you want to connect). Claude will walk you through the authentication process to securely link your accounts.
Step 3: Start asking questions. Instead of logging into HubSpot and clicking through reports, I can now ask Claude directly: “How many contacts and deals do I have in HubSpot?” and it analyzes my data, counts deal stage occurrences, and gives me a summary like “Contacts: 2,691 total” along with insights about data discrepancies.
The real magic happens when you realize you’re having a conversation with your business data rather than wrestling with dashboards and exports. What used to require multiple clicks, filters, and manual analysis now happens in plain English. This is the true future of business intelligence.
Connectors are only the beginning. Tools for vibe coding, such as CoPilot with Visual Studio Code, allow you to switch the state of your AI instances from ask, which is the traditional chatbot structure of question and answer, to agent. As an agent, it’s possible for Claude and ChatGPT to not only read records like a list of contacts, but also act on those records and clean up the format of phone numbers or move legacy data into a new form field.
ChatGPT 5 now includes a toggle that allows you to switch the state from this ask state to the agent state.
This year at Microsoft Build, Microsoft began showcasing how it’s deeply integrating MCP services into Windows, allowing agents to perform tasks through standard applications like Paint or apps that support the technology, like Slack and Figma. Microsoft’s approach also allows you to do this with a local LLM from a repository of AI models like Hugging Face, which gives you the option to train and customize, and keep all of the data locally on your machine.
Once you start having natural language conversations with your business data, whether it’s asking about your biggest deals, identifying bottlenecks in your sales process, or getting instant updates on project status, there’s no going back. Organizations that don’t structure their data and systems to enable these kinds of interactions will find themselves at a serious disadvantage.
If you’re looking to quantify this for your management team or the bean counters, I have a few calculators I’m working on that you should check out. One looks at the amount of revenue you can potentially unlock, and the other measures efficiency by showing the productivity boost an organization might achieve as hard dollars.
-jason
📼 Talking Tech: Watch & Learn
🔗 The Best in Tech This Week
🕶️ Meta’s Smart Glasses Finally Get a Screen – Meta is back at the wearables table with smart glasses that actually have a display and a wristband controller. I’m excited to see how they feel in real use.
🔋 iPhone Air’s Real Breakthrough: The Battery – Everyone wants to talk about how thin the iPhone Air is, but the real story is Apple’s new battery tech. I’m still debating whether to go big with the iPhone 17 Pro or try the ultra-thin Air, which feels like a preview of where Apple could take its hardware next.
🖥️ Microsoft Turns to Anthropic for Office – Microsoft reportedly tapped Anthropic’s Claude after it outperformed OpenAI in certain Office tasks. If true, it shows just how fragile alliances are in the AI race, and that even OpenAI’s biggest backer won’t hesitate to swap engines if someone else gets better at Word and Excel.
🎤 The AI Roadshow: Workshops, Talks & Beyond
Sept. 25 – BannerX w/ The Baltimore Banner
Sept. 26 – Atlanta University Center Data Science Symposium
Oct. 3 – DevFestDC 25
Nov. 04 – WTCI AGILE Series
Nov. 20 – AI Summit Europe
📕The AI Evolution
I wrote The AI Evolution as a practical guide for leaders, builders, and anyone interested in learning how to use AI effectively. This book is about clarity, strategy, and what it takes to bring AI into your organization.
If you’re an executive trying to shape AI strategy, a manager looking to empower your team, or a developer wondering how this shift will change your craft, this book was written with you in mind.
P.S. Before you go…
Zuckerberg vs Zuckerberg is what happens when a platform has to find that line between protecting its founder and realizing that Mark is not the only Mark Zuckerberg on the planet. A bankruptcy lawyer from Indiana has been practicing law since the Meta founder was three years old, but Facebook’s automated systems keep flagging him for ‘impersonating’ himself – disabling his business page despite spending over $11,000 on ads. He’s even created iammarkzuckerberg.com to chronicle how sharing a name with a tech billionaire has turned his life upside down, from prank call assumptions to limo chaos at speaking events.
