An AI agent named David sits in on your SAP workshops, tracks the scope, and writes the test scripts - so the rollout stops feeling like a two-year project.
There is a number that everyone in enterprise software knows and no one likes to say out loud: more than 75% of implementation projects run over budget, over time, or both. SAP rollouts, Salesforce migrations, the multi-year transformations that big companies stake their operations on - a comfortable majority of them go sideways. This is not a secret. It is closer to a law of nature. And laws of nature are exactly the kind of thing a startup likes to argue with.
Luzid is that startup. Founded in 2025 and based in San Francisco, it makes an AI agent called David. The name is a small joke with a large point: one modest copilot, taking on the Goliath of enterprise implementation. David does not try to replace the consultants in the room. It tries to remember everything they say.
That distinction matters, and Luzid leans on it. "David is not an automation tool," the company says. "It's a copilot for software implementations." The difference is not marketing. Automation tools are brittle - they do one scripted thing and break when reality shifts, which in an SAP project is roughly hourly. A copilot is supposed to adapt: to listen to a requirements workshop, notice that a decision made in week three contradicts an assumption from week one, and flag it before it becomes a change request with a price tag attached.
The mechanics are organized into three pillars, which is a tidy way of saying David shows up for the whole lifecycle. There is Workshop Intelligence, where the agent listens to implementation workshops, prompts consultants with best-practice questions, and captures every requirement and decision so nothing evaporates between the whiteboard and the go-live. There is Project Intelligence, which watches scope and change requests continuously and surfaces problems while they are still cheap. And there is Testing Intelligence, which auto-generates test scripts and keeps the evidence traceable - the part of the job that quietly eats weeks and that nobody volunteers for.
The results Luzid reports are the kind of numbers that make an implementation partner lean forward: up to a 90% reduction in the effort of designing business blueprints, roughly 70% off the time it takes to create test scripts, documentation effort cut by 60 to 90%, and testing cycles compressed by one to two weeks. Numen, an SAP Gold Partner, is the named reference customer, and it reports cutting the time spent on test-script generation and integrated testing by up to 70%. The company describes all of this as a 10x acceleration of key deliverables, which is the sort of claim that invites a raised eyebrow - though the underlying idea, that AI is unusually good at the documentation-and-testing grind, is not controversial.
The most interesting design choice is the one that is hardest to demo: memory. Every SAP go-live tends to start from zero. New team, fresh mistakes, lessons that walked out the door with the last set of consultants. David is built to retain context across projects - to absorb the knowledge generated at every step and compound it over time. The pitch is that the second rollout should be smarter than the first, and the tenth smarter still. If it works, the moat is not the model. It is the accumulated institutional memory that a rival would have to rebuild from scratch.
Success means that software implementations stop feeling like 'projects' and start feeling like simple software updates.
David sits in on implementation workshops, prompts consultants with best-practice questions, and captures every requirement and decision - so nothing is lost between the discussion and the delivery.
Continuously monitors scope, change requests and best practices, surfacing risk and scope creep before it escalates into a budget overrun instead of after.
Auto-generates test scripts and keeps traceable evidence, compressing test preparation from weeks to days while keeping the whole thing audit-ready.
Accelerate SAP and enterprise software transformation through agentic AI orchestration - raising delivery quality and cutting risk across the implementation lifecycle, so rollouts feel less like campaigns and more like updates.
A universal AI copilot for enterprise software implementations across every major platform, making complex change seamless and process-driven rather than a leap of faith.
Leads product vision and strategy. A Stanford dropout who did deep-learning research at NASA on exoplanet discovery and was, by his own account, the youngest quantitative trader in Two Sigma's history. Research interests span mechanistic interpretability and time-series forecasting.
Brazilian engineer who worked at Meta AI and was a founding engineer at Orby AI before co-founding Luzid. The founding team carries experience from NASA, Meta AI, Microsoft Copilot and Bain & Company.
Rather than chase enterprises directly, Luzid went where the implementation pain concentrates: the SAP and Salesforce system integrators who live inside these projects. Named partners and users include -
Announced a $3M pre-seed round led by NXTP with Oceans Ventures and Quiet Capital, plus angels from OpenAI, Nubank, Auth0 and Isaac.
Published results showing David accelerates test-script generation by up to 10x and compresses testing cycles by one to two weeks.
Became an SAP Build Partner and began working with SAP system integrators including Numen, an SAP Gold Partner.
The product is called David - a lone copilot squaring off against the Goliath of enterprise implementation.
CEO Andres Carranza did deep-learning research at NASA on exoplanet discovery and was reportedly Two Sigma's youngest-ever quant trader.
Luzid was formerly known as Luzidos - the CEO's email still runs on the luzidos.com domain.
Former SAP executives - including a former Global CRO and regional presidents - have publicly recommended the company.
See the David agent handle workshops, scope tracking and test generation across the SAP lifecycle.
The CEO on why a copilot beats automation, and what success looks like for enterprise implementations.