Thinking Outside the Chat Box: How Nick Pericle Puts AI to Work in B2B
November 20, 2026
Operations Analyst & Chief of Staff · The B2B eCommerce Association
Most B2B teams still use AI the way they first met it: type a question into a chat window, get an answer back. In the second episode of Open Laptop Sessions, Nick Pericle of Tenexity showed host Justin King what happens when you move past that.
Nick helps distributors and manufacturers identify, prioritize, implement, govern, and measure AI initiatives across their business. He is also one of the most advanced everyday AI users in the B2B space. In this session he shared his screen and walked through the agents, connectors and tools he relies on, with a steady focus on what manufacturers and distributors can apply now.
He opened with a hypothesis worth keeping in mind. In his experience, what gets discussed in Silicon Valley about how AI is being applied tends to reach B2B distribution and manufacturing about a year later.
For manufacturers and distributors, that lag is both a risk and an opportunity. The tools Nick uses are available today, and the teams that learn them now will be ahead of their competitors when the rest of the industry catches up.
What actually makes something an agent
Nick draws a clear line. To him, an agent can access other systems, make a plan, and carry out repeated steps on its own. A custom GPT that simply saves an extra set of instructions is closer to an assistant. He isn't precious about the vocabulary, but the distinction matters because agents can do work for you once you set the guardrails for how they should act.
To show what that looks like, he signed up for MetaMuse, Meta's new consumer agent, and connected it to his Google Calendar and Gmail live on the call. He asked it to schedule a call with Justin, change the length, and add Justin to the invite by finding his email address. Before sending the invite, MetaMuse asked for permission to notify the guest, which Nick pointed to as an example of how these tools put a layer of control between access and action.
Nick expects MetaMuse to be many people's first real experience with an agent, simply because so many people already use Facebook and Instagram. For anyone who has only used AI to plan vacations or write emails, he predicts it will be an "awakening moment."
Break work into jobs to be done
Nick also showed how he works in Grokbot, where you build a team of agents. He talks to one agent, which he calls his "Master Chief" chief of staff. That agent doesn't do everything itself. It hands work to specialists: one for client follow-up, one for finding emails, one he calls the "over-communicator" to make sure no client is left wondering, and one that turns his ideas into content for LinkedIn and other channels.
As Nick put it, this is how organizations have always scaled. You break work into discrete areas and give each one focused context.
Justin connected it to the jobs-to-be-done idea. A person in a single role might handle sixteen different jobs, and agents work best when you separate them. His advice for getting started is to pick one small job and try it. Nick agreed: test it, learn from it, and expect some nuance along the way.
Consider an inside sales rep at a distributor. In a single day they might answer product questions, build quotes, chase follow-ups, check stock and update the CRM. Each of those is its own job, and each is a candidate for its own agent. It's the approach Richard Enriquez of White River Hardwoods took in the first episode of Open Laptop Sessions, where he built six AI agents to answer customer calls around the clock.
Bring your most ambitious idea
The strongest story in the episode came from a workshop Nick ran earlier that week. After a morning strategy session on AI's capabilities, the group went hands-on with Claude, the tool the company planned to roll out.
One participant was a director of continuous improvement. His job is to analyze situations, crunch data, build models and make recommendations, mostly in Excel and a third-party tool. Nick asked him for the most ambitious idea he had. While he explained it, Nick recorded it into Claude with the microphone button, then hit go.
Claude ran for about 10 minutes. Afterward, the director sat staring at his screen for another 10. When Nick asked, "Did we just break your brain?" the director answered, "I had no idea I could do this."
Nick's takeaway is that when you push Claude and bring real ambition, Claude will be ambitious right back.
If you yourself are not pushing the tools to its limits, you're artificially setting your ceiling too low.
In his view, the hardest part of AI right now isn't enterprise architecture or security. It's imagining what's possible and setting expectations high enough.
Justin added a note for anyone who feels stuck. It took him weeks before AI clicked. His advice is to sit with the barrier for a while, because the breakthrough on the other side is worth it.
AI is no longer an intern
Justin shared a tool he built that scans the job postings of major distributors and manufacturers for AI and machine learning roles, a signal of where those companies are heading. He started with a structured prompt from a prompt-builder project, then used Codex to turn it into scripts that now run weekly and post findings to Slack. He estimates it would have taken him a week to write as a developer. It took about 10 minutes.
Nick used the example to reframe how B2B professionals should think about AI. People often compare it to a really smart intern, but he says that was true about two years ago.
AI is now an extremely capable software engineer.
Most knowledge workers, especially in distribution and manufacturing, haven't yet learned how to direct one. Justin's point: without AI, he would have had to hire someone, or he simply wouldn't have built it.
Nick shared his own example. Working on an ERP integration with an extremely long API document, he explained the task to Codex. In one shot, and in about 30 minutes, it wrote the initial integration to pull data out, built an audit log and retry logic, and created an interface for the client.
For manufacturers and distributors, ERP integration is often where digital projects slow down, waiting on IT backlogs or outside vendors. Nick was clear that internal technical teams and vendors still play an important role, arguably more than ever. But the options for getting started have expanded.
He also compared the two tools. Codex is very good at doing exactly what you ask. Claude is better at taking an idea and turning it into something polished, thinking through areas you didn't specify.
So what about ROI? When a CFO asks, Nick's answer is that leaders need to raise their expectations of what roles and tasks can now accomplish.
You need to expect more out of your people, because of the tools now available to them.
Connect AI to the systems you already use
Justin said he rarely leaves his AI tools during the day anymore, even to search his email. Nick finds himself working the same way more and more, though he admits old habits die hard. He compared it to working with a colleague: you build trust by spending time together and learning their strengths and weaknesses.
Nick showed the connectors he has set up in Claude, including Asana for project management, his CRM, Canva, Fireflies and Granola for meeting transcripts, and Mercury for finances. Instead of logging into each tool, he asks Claude to pull from them as he works.
His most practical example was meetings. He asked Claude how many times AI governance came up across all his meetings in the past week, and what the top use cases discussed were, without naming any clients. Claude pulled transcripts from both Fireflies and Granola, merged them, and answered across the full week.
Both Nick and Justin strongly recommend recording your meetings, whether through Teams, Google Meet or a dedicated note taker. Justin calls it one of the most valuable disciplines he has built, and says it lets him be more present because he doesn't have to take notes. He also pushed back on the idea that AI makes people think less. "I felt like my brain would shrink with AI, when in reality my brain has exploded," he said. The difference, he explained, is using it intentionally.
If your company runs on Copilot, check this first
Justin sees most organizations adopt Microsoft Copilot first, then Claude, then ChatGPT. Nick expects Microsoft to keep improving, but believes Copilot is currently behind the leading tools.
His practical advice: check whether you have the basic or premium version of Copilot. If basic has been your only experience with AI, he would bet it has been underwhelming. Ask for premium access, even for 30 days. He compared the gap to ordering a lemonade and getting half lemonade, half water.
A factory floor from a prompt
Nick closed with something visual. Using Claude with the Higgsfield integration, he took a few photos and a description of a manufacturing facility and generated a 3D model he could zoom in on, including where cameras would be placed. Work like that would once have required a significant investment. He did it with a prompt.
Justin saw the potential right away, imagining what this could mean for bills of materials and visualizing parts in real time.
Where to start
The through line of the conversation is that the tools are already capable of far more than most B2B teams ask of them. A few starting points for manufacturers and distributors, drawn from the episode:
Get hands-on with an agent, not just a chatbot, and connect it to one system you use every day.
Pick one small job to be done and hand it to AI before trying to automate a whole role.
Bring your most ambitious idea and see how far the tools go.
Start recording your meetings so you can query them later.
If you're on basic Copilot, ask for premium.
Open Laptop Sessions features candid conversations with B2B leaders and practitioners showing what they're actually building with AI. Laptops open, real systems, no slides. For more conversations like this, join us in Scottsdale for B2B NEXT AI 2027, March 1-2.