The short version
David Friedberg described how his company assembled an internal CRM over a weekend using Claude Code, Cursor, and related tooling. The application ran and the team used it. That initial success did not settle the build-or-buy decision.
Once the CRM became operational, the scope expanded: new features, account management, security, access control, protected data storage, and the work required to make the system useful at scale. At the same time, the same team had opportunities to build plant-breeding software and laboratory workflows that were unique to the business.
The company ultimately adopted Salesforce. Friedberg’s stated lesson was not that AI coding failed. It was that internal development time produced a better return when spent on workflows that created a distinctive advantage rather than recreating horizontal software.
Who is David Friedberg?
David “Dave” Friedberg is a technology entrepreneur whose work has repeatedly connected software with agriculture and life sciences. He is CEO, chairman, and co-founder of Ohalo Genetics, the agricultural technology company discussed in this case. Ohalo develops plant-breeding technologies intended to improve crop productivity.
Friedberg also founded and leads The Production Board, a venture foundry that builds and invests in businesses across food, agriculture, and life sciences. Earlier, he founded The Climate Corporation, a digital agronomy platform acquired by Monsanto in 2013. Before that, he worked in corporate development and product at Google and studied astrophysics at the University of California, Berkeley.
He is also one of the four hosts of the All-In Podcast, where his comments about software economics and the “SaaSpocalypse” thesis helped bring this case wider attention.
That background is relevant to the analysis. Friedberg is not describing a hypothetical startup exercise: he leads a company that develops proprietary scientific technology while also deciding which horizontal business systems it should operate. His account therefore illustrates the allocation problem at the center of build versus buy—where a capable technical organization should spend its limited development time.
What the source establishes
| Claim | Evidence status |
|---|---|
| The company built an internal CRM with Claude Code and Cursor | Stated directly by Friedberg |
| The initial version was assembled over a weekend | Stated directly by Friedberg |
| The team used the internal system | Stated directly by Friedberg |
| Feature, security, access, and data requirements expanded the scope | Stated directly by Friedberg |
| Development competed with proprietary plant-breeding software | Stated directly by Friedberg |
| The company subsequently adopted Salesforce | Stated directly by Friedberg |
The segment does not disclose the team size, exact engineering hours, Salesforce plan, contract value, migration cost, maintenance cost, duration of production use, or a numerical return-on-investment calculation. Code or Buy therefore does not infer or publish those numbers.
The build was real—and still lost
The most useful part of this case is that the prototype appears to have worked. This was not a story about an AI tool failing to generate a database and interface. The reversal came after the first usable version.
That distinction matters. A prototype answers, “Can we make the workflow run?” An operational system must also answer:
- Who is allowed to see and change each record?
- How are identities, roles, and departing employees handled?
- Where is sensitive customer data stored and protected?
- What happens when records are duplicated, imported incorrectly, or deleted?
- Which integrations must remain reliable?
- Who responds when the system fails during a critical sales process?
- How does the system evolve without consuming the team’s roadmap?
AI-assisted coding reduces the effort required to create features. It does not eliminate ownership of the resulting system.
The decisive cost was opportunity cost
A conventional comparison might place a Salesforce subscription on one side and engineering hours on the other. Friedberg’s explanation introduces the more important comparison: CRM development versus proprietary product development.
His company had scientific and plant-breeding workflows that general-purpose vendors could not reproduce as effectively. Every hour spent improving permissions or account behavior in an internal CRM was an hour unavailable for software that could strengthen the company’s core capability.
The relevant question was therefore not simply whether the CRM could be built for less than the license price. It was whether CRM was the highest-value use of the team’s limited technical capacity.
Code or Buy assessment
Buy the horizontal system of record
CRM is an established horizontal category. Mature products already contain years of accumulated work in permissions, audit history, integrations, import and export, reporting, mobile access, reliability, and administration. A company adopting one of these products is purchasing operational maturity as well as visible features.
For a business that needs CRM to function reliably but does not compete on CRM, buying is usually the stronger default.
Build the vertical advantage
Software for a proprietary breeding process has different economics. The workflow may encode company-specific knowledge, improve research throughput, and compound into an advantage that competitors cannot purchase from the same vendor.
This is where AI-assisted custom development can have its highest strategic return: not cloning every feature of a horizontal SaaS product, but making unique processes easier to execute.
Use a hybrid architecture
The resulting recommendation is HYBRID:
- Keep customer and account data in a professionally operated system of record.
- Build the scientific, operational, or product-specific workflows that create differentiation.
- Connect the two through supported APIs instead of rebuilding the entire horizontal platform.
- Revisit the boundary if license cost, data portability, or workflow constraints become materially worse.
When the verdict could change
Building a CRM can still make sense when the required system is narrow, the data model is simple, the number of users is small, and failure has limited consequences. It can also make sense when conventional CRM assumptions fundamentally conflict with the workflow.
The build case becomes stronger when:
- per-seat pricing grows far faster than maintenance cost;
- the organization already owns identity, audit, security, and integration infrastructure;
- CRM behavior is itself part of the product or competitive advantage;
- an existing system cannot support the required data model;
- the team can commit to owning the system for several years.
The buy case becomes stronger when the CRM is a canonical source of truth, many teams depend on it, compliance matters, or integrations are numerous.
What this case does not prove
This account does not establish that Salesforce is the best CRM for every company, or that internal CRM projects are inherently wasteful. It also does not provide enough financial data to show which option had the lower accounting cost.
It demonstrates a narrower and more useful point: fast implementation can make a build look inexpensive before the cost of operating the system and the value of displaced work are understood.
Sources and editorial method
The primary source is Friedberg’s first-person conversation with Salesforce CEO Marc Benioff during Salesforce’s FY27 second-quarter earnings broadcast on 26 August 2026. Benioff introduces Friedberg at 24:17; the weekend CRM account begins at 25:23; and their discussion continues to approximately 33:12.
The setting matters: Salesforce produced the broadcast, Benioff conducted the interview, and Ohalo had become a Salesforce customer. It is therefore both direct evidence of what Friedberg said and vendor-controlled customer marketing. Code or Buy treats Friedberg’s description as a first-person account, not independent proof of Salesforce’s comparative performance.
- Watch the official Salesforce earnings broadcast from Friedberg’s introduction
- Jump to the weekend CRM account
- Read a timestamped earnings-call transcript
- Read Friedberg’s official biography on All-In
- See Ohalo’s leadership page
- Listen to the later All-In discussion
Code or Buy independently summarized and analyzed the public material. We did not interview Friedberg, inspect the internal CRM, or verify undisclosed costs. Friedberg, Ohalo, Salesforce, Anthropic, and Cursor did not review or sponsor this report.