Larridin Raises $30 Million in Series A Funding: How Should Enterprise AI Calculate Its Return on Investment?

tradeNovember 4, 2026release AiFun

Abstract: Larridin has secured a new round of funding to link the use, costs, and outcomes of enterprise AI. The need it addresses is very specific: simply purchasing tools and increasing their usage does not directly answer the question of what value the enterprise actually gains. A truly useful metrics system should take into account the quality of completed work, manual review, and business outcomes, while also acknowledging that different metrics cannot be easily interchanged.

A Funding Round Highlights the Need for Accountability in AI Budgets

On November 2, 2026, Larridin announced the completion of a $30 million Series A funding round led by Pelion Ventures, with participation from existing investors including a16z and Gradient Ventures, bringing its total funding to $47 million. The company positions itself as an enterprise AI value measurement platform that links AI adoption, output, and spending to the work of employees and AI agents, helping businesses determine where to allocate their next investments.Larridin Funding Announcement

The significance of this initiative lies in its attempt to address the lack of post-purchase accountability. Budget managers may know how many seats were purchased and how much was spent on model calls, but they may not be able to link these expenses to specific outcomes. If, when renewing subscriptions, scaling up, or replacing tools, the only data available is the number of logins, the basis for decision-making is very weak.

However, the platform itself must also be held to the same standard: Can it help organizations make better decisions, and do the benefits outweigh the costs of accessing, maintaining, and interpreting the data? Securing funding indicates that investors are willing to support the business’s continued growth, but it does not, on its own, prove that the product has established a universally applicable standard for calculating returns.

Usage Rates: Delivery Volume and Business Value Should Be Considered Separately

Larridin’s third-quarter engineering AI report shows that the proportion of AI-assisted merged code change requests at its sample companies rose from 35% in April 2026 to 62% in September. “AI-assisted” includes both direct and inferred attribution; this percentage does not represent the share of work completed independently by AI, nor does it represent the entire industry.Larridin Report

This definition is important. While “62%” indicates an increased level of AI involvement, it cannot be directly interpreted to mean that “the 62% code was written entirely by AI,” nor can it be directly converted into a specific amount of labor cost savings. There are still several factors that must be considered between the degree of AI adoption, engineering output, and business outcomes—such as whether the requirements are correct, whether the quality meets standards, and whether the results are actually being used.

For example, if a team completes an internal feature faster but customers don’t use it, engineering throughput may still increase, yet business revenue may not necessarily rise. Conversely, a very small fix can be highly valuable if it resolves a critical issue. These are merely hypothetical scenarios illustrating differences in metrics, and they serve as a reminder that companies should not let easily quantifiable numbers replace the outcomes they truly wish to improve.

An increase in output and a rise in unit costs can occur simultaneously.

The report also shows that per-engineer output increased by 33% from the second quarter to the third quarter; AI spending per unit of output was 89% higher in September than in May. The latter average was influenced by a company that adopted AI later than others; for some companies, the September level was lower than that of May.Larridin Report

These two sets of figures have different time frames and measurement criteria, so they cannot be compared to draw a unified conclusion about efficiency. Together, they serve as a reminder to managers that when they see an increase in output, they should also examine the rate at which costs are changing.

From a management perspective, this discrepancy is precisely what should be preserved. A company may be willing to pay higher costs for faster delivery, or it may find that additional resources do not yield corresponding benefits. To determine whether this is a good or bad outcome, it is necessary to know whether the current objective is to shorten delivery time, reduce costs, or improve quality. If the objective is unclear, the same report can easily be used to support opposing conclusions.

In an earlier public benchmarking statement, Larridin also acknowledged that the relationship between expenditure and output is correlational and does not prove causation; a metric centered on merged code tends to underestimate the output of personnel primarily responsible for review, operations, and maintenance.Larridin: Public Data and Methodology Description

How can companies establish their own calculation methods?

A more practical starting point is to first identify complete, recurring workflows rather than immediately ranking individuals company-wide. For example, choose the entire process—from receiving a customer service request to resolving the issue, or from submitting an engineering request to completing the delivery—and then define what constitutes a successful completion.

Next, list all the costs in full. In addition to training and model invocation, you should also factor in integration and maintenance, manual review, error correction, and the impact of wait times in the process. If you only track the generation phase, AI might speed up a particular step but shift more work onto downstream reviewers. In this case, while individual metrics may look good, the overall process may not necessarily improve.

When making comparisons, it’s important to consider the context: Have there been changes in task difficulty, staffing, tool versions, or the scope of the metrics? Prioritize comparing similar tasks, continuously monitor quality and completion times, and then decide whether to scale up investment. Time savings, in particular, require careful explanation. They may represent new capacity, or they may simply be spare time scattered across various tasks; only when expenses are actually reduced can they be recorded as cash savings.

Finally, assign a specific purpose to each metric. Cost anomalies are used to review billing and usage patterns; delivery cycles are used to identify bottlenecks; and quality metrics are used to assess risk. If these are aggregated into a single overall score, users may not see the trade-offs behind the improvement in that score.

Even products used to measure work have their own limitations.

For platforms like Larridin, it is worth further examining whether metrics from different business scenarios can be interpreted and verified, and whether the recommendations have actually influenced corporate decision-making. Just because engineering code makes it easy to maintain a record does not mean that the same approach can be directly applied to sales, research, or management.

Before implementation, companies should also clearly define the scope of data collection, access permissions, and intended uses, and ensure that employees understand how this information will be used to improve specific processes. In particular, when metrics have blind spots, they should not be used as the sole basis for evaluating individual performance. This is a recommendation regarding implementation practices and does not imply that Larridin has such issues.

My view is that the value of measuring the return on AI investment lies in enabling organizations to continuously refine their procurement and work arrangements, rather than in providing a single, permanently valid percentage. Larridin’s funding has secured more resources for this approach; how far it can go depends on whether it can further transform “seeing the data” into evidence-based choices that stand up to business scrutiny.