It’s a common assumption in the business world: Legal, compliance, and risk departments are the last to change their ways. Historically seen as risk-averse, slow to adopt new technology and protective of established processes, these functions are labeled as often lagging when it comes to enterprise transformation journeys.
At Salesforce, we decided to test that assumption and turn it on its head. And in testing it, we may have built something more useful than expected: a replicable playbook for how any organization, in any industry, can responsibly embrace AI and agentic transformation at scale.
Salesforce Legal and Corporate Affairs (LCA) is a global team that manages hundreds of thousands of requests per year, and this transformation has fundamentally changed (and continues to change) how we operate, govern, and partner with our colleagues across the company.
As Customer Zero for our own technology, LCA is building an AI-native, trust-first, agentic legal function that holds the promise of defining the best of human-AI collaboration. Humans and agents work together to orchestrate agentic workflows. Agents handle routine questions, freeing our attorneys and other legal professionals to focus on what only humans can do: building trust, leading with integrity, and solving high-stakes, complex problems. Team members develop their careers by building agentic fluency, AI skills, and cutting-edge prowess.
Beyond productivity, one deeper value is commercial velocity. When legal response time accelerates, when sellers are empowered through effective self-serve delivery models, and when teams are augmented with agentic capabilities, we remove friction, simplify processes, and close deals faster. At scale, this compounds, serving as a true force multiplier for our customers, partners, and business.
Here’s part of the playbook we’re building in real time.
Slackbot powers our Salesforce legal team
The integrated AI companion solves a critical friction point: the need for high-precision internal research, trusted information finding, and document and thread summarization without disruptive context switching. Because Slackbot leverages curated source-of-truth internal knowledge articles and legal resource libraries built by trusted teams, it delivers accurate, consistent guidance the moment our people need it. It’s a real-time intelligence layer that keeps our global team aligned, accurate, and agile.
Real-world legal use cases
Here’s how Slackbot helps our global legal team move quickly, with trust at the forefront:
1. We don’t have to search for our data now. We can just talk to it. In the past, checking the status of a main service agreement (MSA) or finding a specific business associate agreement (BAA) contract meant switching between multiple Salesforce instances and databases. By design, Slack and Slackbot work seamlessly alongside core CRM systems and records, including Salesforce, to streamline workflows. Now, my team asks questions in plain English directly in Slackbot, like:
- “What’s the status of the MSA negotiation with [Company X]?”
- “Give me a summary of the top three themes in our data privacy requests this year.”
Slackbot pulls the answer from our “source of truth” and provides a direct link to the record. No context-switching, no hunting for tabs.
2. We have streamlined our verification processes. The legal team has built a product legal front door in Slack, a self-serve Slackbot-powered intake experience that helps Technology and Products colleagues get to the right legal resource quickly. Depending on the query, Slackbot can route users to existing legal channels, prompt them to submit a customer case, determine whether a Product Legal Review (PLR) is needed for a new feature or product, or walk them through an intake process for new product partnerships. This is anchored in trusted, structured legal knowledge articles and reliable data, allowing teams to move quickly with confidence. Looking ahead, the team plans to expand self-service capabilities to cover topics like end-of-life processes and nondisclosure agreements (NDAs).
3. We use Slackbot as our digital assistant. Beyond the specialized legal queries, Slackbot has become a daily productivity companion. We use it to handle the administrative tax of our profession, the fabric of every lawyer’s day, so our human teams can focus on high-value strategy and counsel:
- Meeting prep and summarization: We can turn an hour-long sync into a bulleted list of action items.
- Complex document synthesis: We work through 50-page briefs to pull out specific clauses or summarize complex concepts in seconds.
- Asset creation: We generate initial drafts for documents and presentations, allowing our lawyers to start at 80% completion rather than a blank page.
4. We stress-test legal arguments. This is where it gets interesting for practitioners. Our corporate counsel now uses Slackbot to stress-test arguments before they are finalized. By asking the AI to identify logical fallacies or anticipate potential counterarguments, we aren’t replacing the lawyer’s judgment; we’re sharpening it. It’s about entering a negotiation with a higher level of confidence and getting a better understanding of what our customers may need.
Our corporate counsel now uses Slackbot to stress-test arguments before they are finalized
A new role for a new era: The agent manager
As the legal team integrated AI agents, such as Slackbot, into their daily workflows, it became clear we needed a dedicated role at the intersection of legal expertise and agentic risk. Not a blocker, but an enabler with guardrails. Someone who could proactively mitigate legal and reputational risk without becoming a bottleneck for implementation.
So we created a new human role: the agent manager.
Our goal was to build a repeatable engine for “yes, here’s how to do it safely,” rather than “no by default.” Over 18 months of manual testing and risk-spotting, our agent manager built and operationalized:
- A Risk & Impact Scoring System, developed with LCA subject matter experts, to prioritize the testing of high-risk queries and high-impact processes
- SME “tiger teams,” user acceptance testing frameworks that validate agent accuracy before launch
- Standardized guardrails and disclaimers with cross-functional teams
- Automated regression testing to monitor agent performance as models improve, content is updated, and scope increases
- A checkpoint in the AI Council Review Process that triggers the testing framework whenever a use case is grounded in legal content
We’ve made our agent manager role a permanent and core fixture of our team, and every agent deployment they have supported has successfully launched. Having a human sit at the intersection of the legal profession and agentic technology makes agentic deployments successful, with higher trust and adoption within the legal team.
The broader lesson: A playbook for the Agentic Enterprise
Legal departments are a fascinating test case for AI transformation precisely because of their historical resistance to change. The professional culture values deliberation, precedent, and risk avoidance. The work involves privileged information, reputational exposure, and significant stakes. The talent pool is, by training, skeptical.
If AI and agentic transformation can take root and thrive in that environment — and deliver the kind of measurable results, external recognition, and cultural shift we’re seeing — it provides a proof point and a playbook for the entire enterprise.
What we are building is also a vital signal to our customers. Our broader ecosystem can look at what we’ve built in LCA and see not just a proof of concept but a model. When organizations navigate their own AI journeys, grappling with data governance, legal exposure, and responsible use, they need to know we hold ourselves to the highest standards. Every guardrail, risk framework, and testing checkpoint we built reflects how AI should be deployed in any high-stakes environment: with intentionality, human accountability, and trust built in from the start. We aren’t asking our customers to go anywhere we haven’t gone ourselves.
We are not just adopting trusted AI. We are proving how human and AI collaboration can transform the legal profession — and then sharing what we learn.
That is what it means to be Customer Zero.
