Lightfield: Legacy CRM wasn't built to work with agents
CRM software has spent the last two decades helping companies keep track of customers.
Lightfield thinks the next two decades may look rather different.
More AI-native than the likes of Salesforce and therefore potentially appealing to developers and their operations team counterparts who will implement it, Lightfield is building what it describes as a system of record for organisations where humans and AI agents work side-by-side.
The company’s platform automatically captures customer conversations, decisions and relationship history, then turns that information into a shared source of truth that both people and agents can access.
Building a customer world model
The company was founded in late 2025. Its platform builds a world model of every customer interaction for companies — from the calls, emails, and meetings a company already have — with no data entry required. On top of that model sit agents that prep every meeting, work every deal, and find pipeline in the customers a company already has.
Lightfield’s users are typically AI-native companies, where engineers, product managers and go-to-market teams all work from the same customer record.
The platform reflects what CEO Keith Peiris defines as a customer agent platform, moving beyond rigid account, opportunity and contact records to capture customer conversations, commercial value, product usage and other context by default, creating a “corpus of data” about each customer.
Beyond legacy CRM
The company’s technology proposition hinges on a promise to “shift CRM” from a static database into an active, thinking workspace.
“For the last twenty years, software companies were organised around people. Sales teams managed customer relationships, customer success teams managed implementations, product teams gathered feedback. Information moved between departments through meetings, handoffs, dashboards and software designed for human operators,” notes Lightfield CEO Peiris.
AI is of course changing that structure.
Peiris and team remind us that across engineering, product, support, operations and go-to-market functions, agents are beginning to perform work that previously required dedicated teams. Information no longer needs to move slowly between people.
Agents can access it, reason over it and act on it directly. According to Peiris, that shift is also creating a new class of customer-facing technical roles, often referred to as “forward-deployed engineers” – technical specialists embedded directly with customers to integrate software, build integrations and solve custom problems.
According to Peiris, the way CRM is used today (i.e. beyond being a mere sales tool) at reality creates new infrastructure demands. The systems companies rely on today were built for organisations where humans were responsible for capturing, organising and distributing information throughout the business.
They were not designed for organisations where people and agents work together.
“The software stack that defined the SaaS era reflects that reality. Systems like Salesforce became the source of truth because people were responsible for updating records, managing workflows and carrying information between teams,” said Peiris. “They don’t know what actually happened on the call, in the email, in the meeting — and agents working from that record are just guessing faster.”
Peiris says that agents need access to customer history, decisions, conversations, implementation details, product feedback and relationship context. They need systems designed for reasoning and action, not systems designed primarily for manual data entry.
Adding AI features to legacy software does not change the foundation underneath it, so companies adopting agent-driven workflows require a new kind of system of record.
“Lightfield builds its own record of every customer — from the calls, the emails, the meetings — without anyone typing it in. That record is what your people and your agents both work from, so nobody’s guessing and nothing gets lost between teams.,” explained Peiris.
Instead of relying on manually updated CRM records, scattered notes and disconnected systems, organisations gain a continuously evolving source of customer intelligence that reflects what is actually happening across the business.
Peiris: Most companies still have customer knowledge split across teams. An AI-native CRM brings that information together by default.
The goal is to give engineering, product, customer success and sales teams access to the same customer truth, allowing them to understand customer requirements, implementation history and the reasoning behind feature requests from a shared system of record.
Product teams can understand customer requirements. Engineers can ramp up on a customer more quickly, understand what they need, influence the roadmap and get to the root of the problem a customer is actually asking about. Customer-facing teams can retrieve relationship context. Agents can reason across the same information to automate workflows, support decisions and execute tasks.
As organisations adopt new ways of working, Peiris promises us that Lightfield will provide the foundation that allows people and agents to operate from a shared understanding of the customer.
CEO deep dive
The Computer Weekly Developer Network (CWDN) sat down with Lightfield CEO Keith Peiris for a deeper dive into the company’s technology to find out what it really means to use the company’s platform.
CWDN: You argue that AI is creating a new generation of customer-facing technical roles, including forward-deployed engineers. How does that change who a CRM is built for?
Peiris: CRM used to be built mostly for sales teams because they owned the customer relationship. Product, engineering and customer success worked from different systems, and customer information moved through meetings, tickets and handoffs.
That model is changing. Forward-deployed engineers, product teams and AI agents now need direct access to the same customer context. They need to understand the implementation, the commercial relationship, the feature requests and the reasoning behind them. CRM has to become a shared customer system for everyone working with the customer. The buyer is still the founder or the operator running go-to-market — but the users are everyone who touches the customer, engineers included. That’s the shift.”
CWDN: What software system architectural changes are required to build a system of record that both people and AI agents can rely on?
Peiris: Legacy CRMs were designed around structured fields, because humans were the ones typing things in. Agents don’t need fields — they need what actually happened. The calls, the decisions, the history, in the customer’s own words. Enough context to reason and act on, not enough context to fill in a form.
AI has turned this on its head. Agents don’t just need a few structured fields. They need conversations, decisions, implementation history, product usage, commercial context and relationship history. They need enough context to reason, make decisions and take action.
CWDN: How does an agentic CRM change the way engineering, product, sales and customer success teams work with customer information?
Peiris: Most companies still have customer knowledge split across teams. Sales has commercial context, customer success has implementation details, product has feature requests and engineering usually gets the compressed version of the problem.
An AI-native CRM brings that information together by default. Engineers can understand what customers actually need, product can see how often requests come up, and customer-facing teams can work from the same relationship history. And agents can use that same context to prepare meetings, draft follow-ups and move work forward.