How Firm AI is changing the way professional firms operate and compete

Nikolaus Grefe

Generic large language models (LLMs) — even frontier models like Fable 5, GPT 5.6, and Kimi K3 — struggle to make sense of a firm’s complex data model.  

For example, let’s say Claude is connected to your system of record through an MCP server. If you ask it to find all deals with a margin above 18%, it doesn’t know if you’re referring to EBITDA, free cash flow, or gross profit. As a result, it guesses.   

Or suppose you work at a private equity firm and need a report summarizing your team’s engagements across company types like LPs, advisors, and portfolio companies. However, your firm’s data model only has a data object called “interaction,” and a company object with other categories. Generic LLMs would not understand your prompts. They simply lack the context to understand what these data objects mean and how they relate to one another, so it will likely require multiple rounds of prompting to produce the right output. 

Professionals don’t have time for trial and error. Senior partners in particular expect their firm’s AI to provide the same level of accuracy they’d get from their analysts and associates.  

Firm AI, on the other hand, understands a firm’s data, unique terminology, compliance rules, and ways of working. It uses a semantic layer — including the firm’s ontology, nomenclature, and acronyms — to translate what prompts actually mean, so the LLM can produce accurate, reliable, and consistent outputs.

Nikolaus Grefe

Firm AI’s semantic layer makes it possible to build skills that perform tasks to a firm’s specific requirements. Those skills can then be combined into playbooks that encode firm methods into consistent, compliant workflows.  

For instance, you can run a scheduled playbook every day that updates your firm’s system of record and keeps leadership informed. It could check inboxes for new deals, matters, and  company and contact information; create new records with this data; then notify senior partners about the team’s activities.

Nikolaus Grefe

The firms that are well-positioned to scale Firm AI are those that have already built a strong data foundation and infrastructure. That means they have rich, contextual data that’s well-maintained within structured systems of record: systems that capture deal data, relationship intelligence, timekeeping and billing data, and other information in a consistent, organized manner.

With this foundation in place, Firm AI can connect firmwide intelligence and surface meaningful insights that inform critical business decisions.

Firms that are still operating off of spreadsheets and unstructured data need to get the groundwork right before they can derive value from Firm AI.

Nikolaus Grefe

Yes and no. Standardized dashboards and reports will remain critical for running a firm. But with Firm AI, professionals can retrieve critical intelligence and get insights that go beyond what dashboards and reports provide by typing questions or requests in plain language. 

For example, private equity professionals can ask the AI for pipeline conversion rates or a root-cause analysis for successful and unsuccessful deals, then get an accurate response within seconds. 

Or, a partner at a law firm can simply prompt the system about which lawyers booked more than a certain number of billable hours in the last three months, and the AI will quickly return an answer based on firmwide data.  

Interacting with the data is very intuitive. You don’t need to be a product expert to get what you need anymore.

Nikolaus Grefe

Many firms overlook how their technology choices will affect flexibility down the road.

The AI landscape is changing rapidly, and we don’t know what AI model is going to be dominant a year from now. Cost efficiency is becoming a major theme. For this reason, I always advise firms to build their skills and workflows within model-agnostic AI platforms that leave them with the most flexibility.

The alternative creates significant challenges. If a firm builds skills directly on a platform like ChatGPT or Claude, it locks itself into a single LLM provider along with its associated token consumption and long-term cost implications. If that firm decides to change LLM providers because a better or more cost-efficient one emerges, it has to transition all its skills and workflows to the new one, incurring relevant switching costs. 

However, a platform built with Firm AI doesn’t tie firms into one specific LLM. Skills and playbooks can be ported to other models, so firms can retain ownership of their intellectual property regardless of what happens in the future.

Nikolaus Grefe

I see AI evolving from running work within individual applications to becoming the layer that connects a firm’s entire tech stack —not just core systems, but every tool where work happens: email, calendars, data platforms, collaboration tools, Microsoft Copilot, and Claude. This fully connected, operational AI is what we call Firm AI. 

With Firm AI embedded across the tech ecosystem, firms have the agentic infrastructure that AI coworkers need to run core business processes on behalf of professionals. A signal in one application can trigger governed, compliant actions across others, creating true end-to-end workflows enterprise-wide.

The question isn’t whether firms will build these agentic workflows, but which firms will do it first, and for how many processes.  

The firms that lead this transformation will have a fundamentally different operating model where their institutional knowledge and methods are captured, governed, and deployed consistently and continuously. This will give them a sustainable competitive advantage that compounds over time.