THE QUICK READ

The story in brief

Coverage dated 16 October 2026 identifies five startups in NetApp Excellerator’s 15th cohort: LightSprint AI, Wald.ai, Lasso Security, Cogover and QUIPU AI. Their stated focus areas offer a useful starting point for founders investigating the work around enterprise AI adoption, rather than treating AI as one undifferentiated market.

Source starting point: Express Computer: NetApp Excellerator selects five startups for cohort 15 — 16 October 2026 ↗. Full sources and context appear below.

What the new cohort includes

Express Computer’s report of 16 October names LightSprint AI, Wald.ai, Lasso Security, Cogover and QUIPU AI as the five selected startups. It describes a group spanning India, the United States and Israel, with access to mentorship and opportunities to explore proof-of-concept work.

The cohort is a snapshot of one programme’s selections. It is not a census of startup demand or evidence that every company in these categories will find customers. The useful reading is to look at the different tasks represented.

Five companies, several different customer tasks

The announcement coverage describes LightSprint AI in AI-assisted software engineering; Wald.ai in data-loss prevention for AI communications; Lasso Security in security for AI applications and agents; Cogover in workflow automation; and QUIPU AI in enterprise data context. These are descriptions of the companies’ stated areas of work.

Our interpretation is that the group points beyond the model itself. A customer may need help building software, controlling information exposure, coordinating work or finding usable context. Each problem has a different owner, buying process and definition of a useful result.

Find the moment when the work breaks down

A broad question such as whether a company wants more AI invites a broad answer. A more useful conversation follows a recent task. What was the employee trying to complete? Which information was missing? Who had to review the result? Where did the work stop?

Consider a fictional service business preparing replies to customer enquiries. Drafting the text may be straightforward, while finding the correct policy, checking who can see it and handing an unresolved question to the right colleague remain difficult. A founder who only studies draft quality may miss the actual reason the task is slow. The example illustrates a research approach, not the performance of any cohort company.

Identify the owner as well as the user

The person who experiences a problem may not control the budget or approve access to the systems needed to solve it. Map those responsibilities before describing a large addressable market. A useful first conversation can end with an introduction to the person who owns the next decision.

The questions below are an editorial research framework. They are not product claims, security certification criteria or requirements announced by NetApp.

Questions for investigating an enterprise AI problem
Area to investigateRecent event to ask aboutDecision owner to identify
Software workA change that required repeated manual reviewThe person responsible for delivery quality
Information handlingA task delayed by uncertainty over permitted dataThe person who approves that use of information
Workflow coordinationA request that stalled between teamsThe person accountable for the completed process
Knowledge contextAn answer that used outdated or missing informationThe person responsible for the underlying records

Use an accelerator to resolve a specific constraint

NetApp’s official programme site describes technical guidance, mentoring and proof-of-concept support. For any founder evaluating a corporate programme, those offers become easier to assess when matched to a particular need. Is the team missing technical feedback, access to a relevant user or knowledge of an enterprise process?

Ask who provides the support, how access works and what participation requires. A programme name on a presentation does not explain what the startup learned or delivered. Keep selection, a proposed test, a completed test and a customer agreement distinct in your own reporting.

Choose the problem before choosing the category label

A category can help explain what a product does, but it cannot replace evidence of a recurring customer problem. Gather examples, compare existing workarounds and look for contradictions. A team may be frustrated by an activity yet unwilling to change it because migration, training or approval would cost more than the expected improvement.

The practical next step is small: choose one task, speak with the people who recently performed it and establish what would justify a change. That can produce a sharper product brief than attempting to build for every challenge grouped under enterprise AI.

YOUR WORKING NOTES

Investigate one unfinished enterprise task

Use these questions to investigate the idea in your own context.

  1. Choose a recurring task and ask someone to describe the last time it happened.
  2. Record the information, people and approvals needed to complete it.
  3. Identify the person who owns the outcome and the next decision.
  4. Compare the existing workaround with the cost of changing the process.
  5. Define one observation that would strengthen or weaken the case for building a product.

A closer look

Does joining the cohort establish a customer contract?

No. Programme participation, a potential proof of concept and a signed commercial agreement are different outcomes. Describe only the stage supported by evidence.

What should founders take from the mix of selected startups?

Use the focus areas as prompts for customer research. Investigate a concrete task and its owner rather than assuming a programme selection proves demand for your own product.

Sources & editorial note

Published 17 October 2026. Cohort details are attributed to the linked 16 October coverage and announcement reproduction; the official programme site provides general programme context. Product descriptions identify stated focus areas, not independently verified performance. Our customer-research questions and fictional example are editorial analysis. This is an independent, unsponsored article. Selection, a possible proof of concept and a commercial contract are separate outcomes.

See our editorial standards or request a correction. For a useful comparison, read I-Corps and the questions that take research outside the building.. Explore the founder topic guides for connected reading paths.