Every growing company eventually hits the same wall. The sales team lives in one platform, the finance team lives in another, and anyone who needs a straight answer about what actually happened today has to open five tabs and hope the numbers line up. AI tools promised to fix this, but most still ask a business to start over, with new logins, new dashboards, and new habits for every employee to learn. For a lot of teams, that is too much friction to adopt quickly, so the AI sits unused while the real work still happens the old way.

Mitchell Jones is the Founder and CEO of Lava, a platform that places an AI agent inside the business tools a company already runs, including point-of-sale systems, legacy ERPs, and Slack. Before founding Lava, Mitchell worked at Dropbox and Meta and built his first startup, Lendtable, a Y Combinator-backed company that helps people save and invest their money. He started Lava at the beginning of 2025 after noticing how much still had to happen before an average business, not just a well-funded one, could put AI to work.

In this episode of Lead with AI, Dr. Tamara Nall speaks with Mitchell about the rule the Lava team built the whole product around, “no new primitives,” how the platform connects to a company's existing software instead of replacing it, and what it looks like when a solopreneur and a company with hundreds of thousands of employees hand a piece of their workday to the same AI teammate.

No New Primitives

The phrase comes directly from Mitchell. Lava's internal rule is simple: do not make people learn new things. Let them use the tools and the habits they already have, and build the AI around that instead of asking a team to adopt a new system on top of everything else. Mitchell traces the idea back to his own experience after selling his first startup. He started experimenting with AI tools and found them genuinely useful, but still too raw for an average business to actually run on. A coffee shop owner or a construction site manager should be able to use AI the same way a well-funded tech company can, without hiring an engineer first.

Why Do Most AI Agents Get Stuck at the Login Screen?

Mitchell says most AI agents work by opening a real website and clicking around it, the same way a person would, until a security system or a service like Cloudflare flags the behavior and blocks it. Lava takes a different approach. It connects to a company's software through what Mitchell calls connectors, backend channels that call a tool's existing systems directly instead of clicking through a browser. One manufacturing customer runs on tools like Slack alongside legacy aerospace systems such as Epsilon 3 and Duro, programs most people have never heard of. Asked to move records between four systems and notify a team in Slack, a typical agent signs into each site in turn and hunts for the right information on each one. Lava checks that the four systems are already connected, pulls the information in a single call, and sends the Slack message in another. A restaurant owner trying to understand which menu items are selling faces a similar gap, tying together spreadsheets and a point-of-sale tool like Toast instead of simply asking a question and getting an answer.

Mitchell compares the difference to two eras of communication: most agents are sending Morse code, and Lava is on the telephone.

That difference in method is also a difference in speed. Mitchell estimates Lava completes the same task in roughly a tenth of the time.

A Coffee Shop Shopify Store and a Spaceship on a Deadline

Two stories stand out among Lava's early customers. One user, trying Lava for the first time, used it to set up an entire Shopify store, including the storefront and the analytics, while sitting in a coffee shop on a weak Wi-Fi connection. The whole process took about an hour. Mitchell says the user's message back to him credited the team with finding something valuable in the product.

The second story involves a manufacturing company working toward getting a piece of hardware into space within a year. Before Lava, the company was moving part orders between four systems that did not talk to each other, and processing five or six orders took about a week. With Lava connecting those systems, the company began moving dozens of orders a day. Mitchell says the company's response was to ask for more work to hand over, not less.

Built for the Solopreneur Doing Five Jobs at Once

Mitchell points out that the first person doing every job at a new company is usually the founder, covering customer support, design, marketing, and engineering all at once. One early user, a retail entrepreneur building a custom-fit bra brand, asked whether Lava could handle mailers. She connected an account, and Lava ran her entire first print campaign, work she had planned to hire out on a freelance platform.

The same flexibility matters on the other end of the size spectrum. Mitchell says enterprise customers configure Lava to their own compliance needs, choosing which AI models it runs on and approving only the specific tools, like Slack or Gmail, that Lava is allowed to touch. Mitchell says guardrails have to be part of a product's design from day one, not added in afterward, especially once an agent has access to a company's internal systems.

What Happens When Every Team Has an AI Teammate?

Mitchell's five-year goal for Lava is straightforward: every team should have a Lava teammate. He points to restaurants as an example since most do not employ a dedicated data analyst but still need to know which menu items are actually selling well enough to keep on the menu. He also says Lava's ability to choose different AI models for different tasks, rather than defaulting to the most expensive model available for a simple job like writing an email, is part of what keeps the product affordable to run at scale.

Mitchell's boldest prediction looks past Lava entirely. He expects more people to start what he calls “leisure businesses,” pointing to ventures like gaming centers as an example, as AI agents absorb more of the operational work that currently consumes a founder's time. He also expects elder care and healthcare broadly to improve as more people shift into that kind of work. Asked what is overrated in AI right now, his answer is “loops,” the practice of building agents that constantly retrain or correct themselves rather than running from a clear rule file. His pick for what is underrated is simply connecting the AI a person already uses to the tools they already have.

Quick Answers

What is Lava?

Lava is a platform that places an AI agent inside the business tools a company already uses, including Slack, Salesforce, and legacy ERPs, so teams can ask it to complete tasks across those systems without learning new software.

How does Lava connect to a company's existing tools?

Lava uses what the company calls connectors, backend channels that call a tool's existing systems directly instead of clicking through a website the way many other AI agents do.

Who is Lava built for?

Mitchell Jones says Lava serves businesses of every size, from a solo entrepreneur managing every part of a business to enterprise teams running legacy software across thousands of employees.

Does using Lava require engineering or technical setup?

No. Mitchell Jones says a user signs into the accounts they already have, and Lava's agent works from there, including from Slack, WhatsApp, or iMessage once it is connected.

Who is Mitchell Jones?

Mitchell Jones is the Founder and CEO of Lava. Before Lava, he worked at Dropbox and Meta and founded Lendtable, a Y Combinator-backed startup that helps people save and invest their money.

For founders and teams who are tired of learning a new system every time they try to adopt AI, Lava is available now at lava.so. Once connected, it works from inside the tools a business already runs, and from Slack, WhatsApp, or iMessage.

For more conversations with the founders building the next generation of AI, subscribe to Lead with AI on Apple Podcasts, Spotify, YouTube, or wherever you listen.

Follow or Subscribe to Lead with AI Podcast on your favorite platforms

Website: LeadwithAIPodcast.com | Apple Podcasts: Lead-with-AI | Spotify: Lead with AI | YouTube: @LeadwithAIPodcast | Facebook: Lead with AI | Instagram: @LeadwithAIpodcast | TikTok: @LeadwithAIpodcast | Twitter (X): @LeadwithAI

Follow Dr. Tamara Nall

LinkedIn: @TamaraNall | Website: TamaraNall.com | Email: Tamara@LeadwithAIPodcast.com

Follow Mitchell Jones (Founder & CEO, Lava)