Legora removes development bottlenecks with AIForce

Growth without the bottleneck

Legora builds AI-powered legal software that helps lawyers research, review contracts, draft documents, conduct due diligence, and analyze cases faster. As they competed for market share in North America and globally, they needed that same speed across their internal operations.

With headcount set to grow from roughly 300 employees to well over 1,000 by year-end, Legora’s sellers and leaders needed faster ways to research customers, prepare for meetings, review pipeline trends, and update Salesforce records. Traditional technology rollouts could take months, while their one-person go-to-market systems team didn’t have the bandwidth to configure every custom field, layout, and workflow. Legora saw an opportunity to equip employees quickly and securely, remove approval bottlenecks, and help teams handle larger workloads without slowing their momentum.

Building at the speed of AI

To keep pace with the market, Legora flipped the traditional development model by adopting a headless approach, connecting Salesforce with AI assistants through Model Context Protocol (MCP) so employees can use natural language to work with live Salesforce data and build AI-powered capabilities without relying on the Salesforce user interface.

Instead of waiting for a central IT team to build and manage every automation, Legora gives employees direct access to Claude and ChatGPT to create agents, custom skills, and automated prompts. Their systems manager utilizes the Salesforce CLI to programmatically deploy database fields, query Salesforce metadata, debug Apex, and troubleshoot errors, while GTM teams can experiment with use cases like pipeline hygiene audits or automated meeting summaries. When an idea works, teams can share and build on it — helping Legora move promising concepts from experimentation to usable capabilities faster.

“With AIForce, eventually nobody has to log in to Salesforce or click anything,” said Cory Gottlieb, GTM Systems Manager. “They can simply use Claude or ChatGPT to update the records they need in Salesforce, while Salesforce serves as the source of truth that other systems pull from. That way, everyone stays aligned across systems.”

Flexibility across AI models

Legora is developing an LLM-agnostic architecture designed to adapt as AI models rapidly advance. After starting with Claude, they added ChatGPT and are A/B testing both models to understand their strengths and how employees use them to interact with Salesforce data.

Rather than rebuilding custom capabilities every time they switch models, Legora centralizes reusable skills, agents, and components in GitHub. Employees can build once and make those capabilities available across Claude and ChatGPT while keeping the same Salesforce integrations and business logic underneath.

Longer term, Legora wants employees to work with Salesforce primarily through AI assistants within Slack, their collaboration hub. They’re exploring how Claude, ChatGPT, Slackbot, and Agentforce can each play a distinct role as more Salesforce account channels connect to Slack.

Fast innovation, firm guardrails

“Salesforce is the critical system of record,” said David Eckstein, CFO. “We moved to Salesforce consciously because at our scale, we needed that enterprise layer. We needed the permissions, we needed governance. We’ve been through several generations of frontier models since we made that move, and we’ll go through several more. The AI world is rapidly changing, but some things are constant. What can’t change every four months is that system of record: Who did we sell to? On what terms? Who should have access to what? If that layer moves, everything sitting on top of it stops being trustworthy.”

Built-in Salesforce controls help keep experimentation from becoming a free-for-all. Every agent inherits the employee’s native Salesforce permissions, so it can’t access or change data beyond what that person is authorized to use.

Legora adds guardrails for data quality and human oversight. A custom Salesforce dictionary skill defines valid metadata and picklist values, preventing agents from writing unsupported values. Before creating or updating a record, a separate create, read, update, and delete skill shows the proposed changes for human approval and offers an immediate rollback if needed.

Salesforce Shield provides Field Audit Trail and Event Monitoring to track agent behavior, API activity, and historical changes. A custom “last modified by Claude” field identifies headless edits, while a centralized GitHub repository distributes vetted skills across teams — giving employees room to experiment without sacrificing consistency, security, or control.

Innovation that keeps pace

Legora saw time to value almost immediately. Their team configured the MCP server in minutes, and completed the initial configuration in less than two days. Within a week, the broader organization gained access — putting live Salesforce data and headless AI development directly into employees’ hands so they could start experimenting right away.

“Overall, Legora sees AI as a capacity expander,” said Eckstein. “Our clients are actually adding more lawyers because they can now take on more work.”

That speed now carries into how Legora innovates. Employees can create new custom skills and workflows themselves, share what works, and continuously iterate, while the systems team centralizes and standardizes the strongest ideas instead of developing every solution from scratch.

This model helps Legora keep moving while they onboard around 125 new employees every six weeks. Individual usage intensity grew about 67% over three months, proving that deep, highly active adoption is expanding alongside headcount.

With plans to reach well over 1,000 employees by year-end, decentralized innovation and reusable AI skills help Legora move new workflows from idea to execution quickly.