Macroscope runs its go-to-market with agents on Lightfield

Macroscope builds AI code review and engineering intelligence software. Its co-founder Kayvon Beykpour calls it "an enterprise understanding engine" that uses a company's codebase as the source of truth. When Claire Rosenfeld joined about a year ago as the company's first non-engineering hire, the CEO's mandate was to stay lean and "AI everything." That meant scaling nearly the entire commercial operation before adding headcount to run it.
One year in, every customer touchpoint at Macroscope - website visits, installs, support tickets, sales calls - flows into Lightfield, and agents act on all of it.

Life before Lightfield: a sprawl of point solutions and workarounds

Before Lightfield, Macroscope had assembled a variety of tools to approximate the capabilities of an agentic go-to-market.
Their previous CRM was unable to consistently source high-quality data on the signals they rely on to qualify and score accounts. Macroscope relies on understanding how many engineers a company has and what programming languages they work in. Much of this context needs to be pieced together from a prospect's open-source repos or, when those aren't public, by combing through job postings. So, like a lot of operators on pre-AI CRMs, Claire routed the automation work out to Zapier. It called an enrichment service to do the research, then pushed the result back into their CRM.
This introduced a new set of challenges. Building integrations and workflow automations to account for deficiencies in their CRM’s native capabilities led to a sprawl of complex workflows that were brittle and difficult to maintain. For example, ingesting new customer sign-ups from Stripe, enriching data, and triggering the correct downstream actions required a Zapier workflow of close to fifty blocks. "My Zapier stuff was insane," said Claire. "It got really expensive, and I was always running out of credits." And every time she wanted to adjust how they engage with their customers, automations required rigorous editing and debugging.

Ingesting every customer datapoint into a connected record

Claire decided to adopt Lightfield because it natively included everything she was piecing together from her old CRM, Zapier, and third-party enrichment services. With Lightfield, she connected every source of customer interactions into a bespoke data model that reflects how her business actually works. Every touch point from learning about Macroscope, to using the product, to engaging with their sales team, to getting support is represented in their CRM.
Known website visitors are passed from RB2B to Lightfield, so Macroscope has a full view of what pages a buyer visits in the customer journey. When someone installs Macroscope's GitHub app, an automation creates the account, writes a note, and researches where the install came from. And when somebody requests access to their trust center - signaling a high intent buyer - an opportunity is created in Lightfield for the team to reach out.
Product downloads, subscriptions, and usage are ingested into the CRM - giving Lightfield and its agent full context on their self-serve customer base. Agents use this data to spot high-usage customers inside their ICP and trigger sales outreach.
Support inquiries are ingested from Pylon - with tickets linked, Slack threads connected, and issues pushed to Linear for the engineering team to address.
Macroscope has even moved to recording every internal meeting in Lightfield, so conversations about how to close, expand, and support new customers add context to the external interactions they have with them.
“I want the AIs to know everything about our customers, so I can ask any question and get an answer that I trust,” explained Claire.

Configuring Lightfield to understand the business

Claire configured Lightfield to model how her business actually works, and uses Skills and Knowledge to teach the Lightfield agent how to understand their data and do work for her.
She started by building a custom object for signed documents. Every contract lives there as its own record, with the signers attached as contacts and a plain-language summary of the terms she cares about, all linked back to the account. Macroscope's CRM now understands its own legal documents: Claire can ask what a given customer agreed to and get an answer, without opening a file.
Other parts of the business became objects too. Product downloads and subscriptions are their own custom objects, related to accounts, so the agent has full context on the self-serve base. Next on her list is a usage object that captures a weekly, structured read on how each customer uses Macroscope's features, linked to the account, so customer health becomes something the system tracks rather than something she checks by hand.
Claire built custom Skills on Lightfield to teach it how to work with customer data. For example, she created a skill to ingest and interpret data from PostHog, where Macroscope keeps its product and customer data - including how to make the right API calls to get what she needs.
She did the same thing with Macroscope's own product, teaching Lightfield how to call the Macroscope API, further integrating customer product usage into their system of record.

Building automations in days instead of months using natural language

Claire has built thirty automations in Lightfield, and can prototype new ones in minutes.
The fifty-block tree in Zapier that managed new customer sign-ups was rebuilt in Lightfield in less than a day, using natural language. "I can build out all the automation that took me months to build on Zapier in a day," she says, "because I can just do it with natural language prompts." She estimates that it now takes 10x less time to reach a working version of a new automation.

This lets her build and scale more complex automations, and offload work that used to require human judgment to agents. For example, when a company applies to Macroscope's open-source program, an automation parses the application, enriches the applicant's GitHub org, checks it against the star and fork thresholds Claire told it to care about, and decides whether the project is worth taking on before it reaches a person.
Lightfield has also reduced the administrative burden of maintaining automations. Whenever something changes upstream and a run fails, Lightfield notifies Claire, proposes a fix via a ‘debug automation’ skill, and it updates itself.

What’s Next

The commercial team that started as just Claire is now growing around her: two account executives, a growth operations lead, a performance marketer, and an events manager. New hires that join don't inherit manual work. The agents Claire built handle the first pass on every inbound signal, qualifying demand and surfacing where a human is actually needed, which amplifies the impact of each new person they bring on board.
Now, Claire and the Macroscope team are focused on scaling an enterprise sales motion. Previously, their growth came almost entirely through a product-led motion; now, their focus is converting that demand into enterprise contracts. Lightfield uses all the context Claire has assembled in the CRM to build custom target account lists scored against their ICP, source contacts to engage, and enroll them in personalized multichannel sequences. Their vision is to retire Clay and Apollo, and run every single customer interaction through Lightfield.
For Claire, building on Lightfield has become the part of her job she enjoys most. “Lightfield is infinitely useful to my team, and building with it gives me so much joy in my life. It feels like the dream AI, a playground where we can just do whatever and connect whatever with no limits.”