Mithril: AI changes the scale of small businesses

Coordinating specialist teamwork — unDraw Team up illustration

IA IA

One person, eleven laboratories: AI changes the scale of small businesses

One person inside the company, eleven laboratories working with it. Mithril Medicine, whose acquisition Anthropic announced on 27 October 2026, changes the question many business owners ask about AI. The opportunity goes beyond writing faster: it is about coordinating work that would normally stretch a small organisation’s capacity.

Its founder, Cody Tranbarger, used Claude to organise contracts, logistics and experiments with eleven external laboratories. A small business does not acquire eleven sets of specialist skills at once. It becomes better able to draw on the people who have them. That coordination capacity is the part worth examining.

Mithril: one person in-house, a network of specialists outside

According to Anthropic, Mithril was a one-person company working on two rare forms of muscular dystrophy. One programme went from an idea to promising results in mice in less than eighteen months. Those preclinical results mark a research milestone; demonstrating a benefit for patients remains work ahead.

The eleven laboratories do not disappear behind Claude. They supply the equipment, methods, experiments and scientific expertise the work requires. The founder organises their contributions. AI helps keep the documents, requests and dependencies between their activities connected.

Headcount therefore tells only part of the story. A single person in-house can draw on a wider organisation of very human partners. For an SME, the attraction is the ability to lead a more ambitious project without immediately bringing every speciality onto the payroll.

What the 341-page report actually shows

The technical report published alongside the announcement documents one week of coordination involving 113 antibody constructs and five contract research organisations. That documented scope is distinct from the eleven-laboratory network described for Mithril. The report refers to a research team: the constructs had already been designed, ordered and produced before the week under review.

When a 341-page production package arrived, a team member asked Claude to review it and recommend next steps. The model produced a visual synthesis and proposed what should happen to each construct. When the team later set aside three groups on production criteria, Claude carried that decision through the plan, the assay run order and the relevant drafts, reducing the programme from 113 to 89 constructs.

This is the mechanism that matters for a business: a decision only helps if its consequences reach the right people. Removing an option also needs to change the order, the schedule, the working document and the message to the supplier. AI can help trace that chain instead of leaving someone to reconstruct it manually.

People still made the decisions—and sent the emails

During that week, Claude drafted 17 emails; the researchers edited and sent 12. Claude sent no email. Team members supplied incoming information, judged the proposals and approved the correspondence. The report also describes errors, some of which survived review: checking a decision does not always reveal a mistaken assumption buried in its details.

For your business, this changes how you delegate. You can ask AI to prepare a file and update the documents affected by a decision while keeping spending commitments, supplier selection and acceptance of work with the responsible person. Review needs to cover the underlying data and assumptions, not just whether the writing sounds right.

For an SME, start where work passes between specialists

Imagine a small brand preparing a launch with a manufacturer, a designer and a logistics provider. A material change could affect the price, visuals, deadlines and product information. In this example, AI could collect responses, flag an inconsistency and prepare requests for confirmation; the brand chooses the trade-off, and each specialist validates their part.

That kind of handover is a better starting point than a vague ambition to “automate everything”. Look for a recurring file that forces your team to reread several exchanges before work can move forward. Our guide to choosing a first AI project for an SME helps you compare it with other priorities, including costs and acceptance criteria.

Coordination then depends on reliable memory: the latest decision, the correct quote, confirmed constraints and where they came from. Business memory that your team and AI can use avoids starting from scratch on every file. Without it, an assistant can produce a convincing answer based on outdated information.

Grow through coordination with your AI Workers

Mithril points to a practical opportunity: your team’s size does not have to determine, on its own, the size of the projects you can lead. Your ability to choose the right specialists and organise their work becomes crucial. Budgets, partner lead times and specialist expertise remain real constraints; AI addresses the work needed to bring those contributions together.

At SAY, this gives AI Workers a specific job: prepare work, retrieve relevant context and track the consequences of your decisions within a defined scope. A first project can be framed around one recurring file, its sources, permitted actions, the person who approves it and acceptance criteria. Before expanding, you check completed files, necessary corrections and the review workload.

What project could your team take on if coordination no longer depended on one inbox? Talk to SAY about scoping a first AI Worker around your business. SAY Go Big Online.