What Breaks Next
Where enterprise AI breaks next — and where the next markets emerge.
What Breaks Next is Mesa Capital Partners’ market-intelligence platform for identifying those constraints before they become obvious markets.
AI is moving from experimentation into production. That changes the questions that matter. The constraint is increasingly not simply whether a model works. It is whether the infrastructure, data, economics, security, governance and operating model around it can support AI reliably at enterprise scale.
We talk directly with the CIOs, CTOs, CISOs, Chief Data Officers, Chief AI Officers, infrastructure leaders and founders making these decisions. Then we look across those conversations for something more valuable than predictions: repeated problems, changing budgets and unmet enterprise requirements.
Those signals inform our research, the companies we advise, and the markets where Mesa believes new infrastructure businesses can be built.
Interviews
Conversations behind the research, published as we record them.
The Agent Is the New Perimeter →
Matan Bar-Efrat, Rein Security
Recorded at Disrupt: Rein Security’s Matan Bar-Efrat on securing AI agents from the inside, why one policy for every agent fails, and closing the loop from alert to fix.
The AI Bottleneck Nobody Talks About Starts With a Single Crystal →
Cameron Khorsandi, Bucket Labs
Everyone talks about GPUs and power. Cameron Khorsandi of Bucket Labs explains why the real bottleneck may be the crystal growth process that comes before every chip.
When AI Agents Outrun Their Guardrails →
Apostol Vassilev, NIST
NIST researcher Apostol Vassilev on why no guardrail removes all of the risk, who should own AI risk, and why “human in the loop” gets harder at the speed and scale of agents.
Physical AI
Megan Dahlgren, Mesa Capital Partners · publishing soon
Six areas of pressure
What are enterprises actually trying to solve?
Our research currently examines six areas where AI adoption is creating new pressure. They are also the areas where Mesa advises the companies and investors building into them.
Economics & Business Value
Where do the costs or promised returns stop making sense?
Architecture & Delivery
What can't the existing technology foundation reliably support?
Data & Knowledge
Where does enterprise information become unfit for machine-driven decisions?
Security & Trust
What can no longer be adequately protected, controlled or verified?
Governance & Accountability
Who owns an AI-enabled decision, its limits and its consequences?
Workflows & Operating Models
Where does organizational work fail to adapt as AI moves into production?
We don’t assume which constraint matters most. We ask the people making these decisions, test the same questions across independent conversations, and look for the patterns that repeat.
From research to decisions
What Breaks Next isn’t research for research’s sake.
Mesa uses this work to help technology companies and investors answer commercially important questions. That is where research becomes market intelligence.
- Is this an interesting technology — or an urgent enterprise problem?
- Who actually owns the problem?
- Is there budget attached to it?
- What would make an enterprise adopt — or reject — the solution?
- What does the first credible enterprise use case look like?
- Where is an emerging infrastructure problem large enough to create a new market?
Executive Briefings
What is actually changing in AI infrastructure?
For CEOs, CTOs and investors who need a clear view of where enterprise AI is breaking — without waiting for the market to name it.
A What Breaks Next Executive Briefing is a focused session for a leadership team, board or investment committee, drawn from what enterprise technology leaders are telling us directly and shaped around the questions you’re working through.
A briefing covers:
- Which of the six constraints enterprises are feeling most acutely right now
- Where budgets are moving — and which problems are still merely interesting
- What enterprise buyers now require before AI reaches production
- Where new infrastructure markets are beginning to form
For AI infrastructure companies
Test the market before scaling the GTM machine.
Technically strong companies can spend millions building sales and marketing around assumptions that haven’t been adequately tested against enterprise buyers. Mesa works with early-stage AI and infrastructure companies to test those assumptions directly.
Market Validation Sprint
A focused engagement designed to answer questions such as:
- Who is the real enterprise buyer?
- How urgent is the problem?
- How are enterprises solving it today?
- What language do buyers use to describe it?
- Where does budget come from?
- What prevents adoption?
- Which use case creates the strongest initial wedge?
- How should the technology be positioned for the enterprise market?
How the sprint runs — typically 4–6 weeks
- 01
Thesis
What problem does the company believe exists?
- 02
Buyer discovery
Conversations with CIO, CTO, infrastructure and AI leaders — including net-new relationships built for your question.
- 03
Market signal
What's urgent, what's funded, what's merely interesting — and what alternatives buyers use today.
- 04
Positioning
Translate the technology into the enterprise problem and buying language.
- 05
ICP + GTM
Who buys first, why, and what the wedge is.
- 06
Executive readout
A founder, board or investor session: what the market said, and what we recommend you do.
The work combines enterprise conversations, market research, competitive analysis and Mesa’s experience across infrastructure and enterprise technology. The output isn’t another market report. It’s evidence founders can use to make product, positioning and GTM decisions.
Talk to Mesa about a Market Validation SprintFor investors
Understand the enterprise demand behind the technology.
New infrastructure categories often become obvious only after the underlying constraint has become painful. What Breaks Next looks for those constraints earlier.
Mesa works with investors exploring emerging AI infrastructure markets, evaluating enterprise demand and helping portfolio companies translate technically sophisticated products into credible enterprise propositions.
Research can support:
- Investment thesis development
- Enterprise market diligence
- Portfolio company positioning
- Customer and buyer intelligence
- Emerging-market mapping
Participate in the research
We are continuously interviewing senior technology leaders responsible for bringing AI into production.
Participants contribute their perspective to the research and receive access to the resulting findings and decision frameworks. Each launch report ships with a five-question diagnostic, a decision tool, and a startup-opportunity brief.
CIOWho Owns the AI-Enabled Workflow?
+
Who Owns the AI-Enabled Workflow?
- LaunchWhat Breaks Next for the CIO: Who Owns the AI-Enabled Workflow?
- 02The AI Business Case After the Pilot: What Does It Really Cost to Operate?
- 03When AI Meets the Application Portfolio: Integrate, Replace, or Consolidate?
CTOCan Your Architecture Support Agents in Production?
+
Can Your Architecture Support Agents in Production?
- LaunchWhat Breaks Next for the CTO: Can Your Architecture Support Agents in Production?
- 02Who Tests the AI-Generated Change? A Research Agenda for Software Assurance
- 03Build, Buy, or Orchestrate? Enterprise Requirements for the AI Stack
Chief Data OfficerIs Your Data Fit for AI Decisions?
+
Is Your Data Fit for AI Decisions?
- LaunchWhat Breaks Next for the Chief Data Officer: Is Your Data Fit for AI Decisions?
- 02Who Owns the Answer? Lineage and Accountability Across Enterprise Knowledge
- 03From Data Access to Data Use: What Must Change When Machines Become Consumers?
Chief AI OfficerWhat Must Be Proven Before AI Can Act?
+
What Must Be Proven Before AI Can Act?
- LaunchWhat Breaks Next for the Chief AI Officer: What Must Be Proven Before AI Can Act?
- 02Which AI Use Cases Survive Contact with Real Workflows?
- 03Who Runs AI After Launch? Ownership, Evaluation, and Escalation
CISOWho Controls What an AI Agent Can Do?
+
Who Controls What an AI Agent Can Do?
- LaunchWhat Breaks Next for the CISO: Who Controls What an AI Agent Can Do?
- 02Can You Reconstruct the Action? Evidence Requirements for Enterprise AI
- 03Where Does Third-Party AI Extend Your Security Boundary?
Report calendar
Our actual publishing schedule, sorted by date.
| Date | Report |
|---|---|
| Sep 14, 2026 | What Breaks Next: Moving AI from Pilots into Enterprise Workflows(published) |
| Sep 21, 2026 | What Breaks Next for the CIO: Who Owns the AI-Enabled Workflow? |
| Sep 28, 2026 | What Breaks Next for the CTO: Can Your Architecture Support Agents in Production? |
| Oct 5, 2026 | What Breaks Next for the Chief Data Officer: Is Your Data Fit for AI Decisions? |
| Oct 12, 2026 | What Breaks Next for the Chief AI Officer: What Must Be Proven Before AI Can Act? |
| Oct 19, 2026 | What Breaks Next for the CISO: Who Controls What an AI Agent Can Do? |
| Oct 26, 2026 | The AI Business Case After the Pilot: What Does It Really Cost to Operate? |
| Nov 2, 2026 | Who Tests the AI-Generated Change? A Research Agenda for Software Assurance |
| Nov 9, 2026 | Who Owns the Answer? Lineage and Accountability Across Enterprise Knowledge |
| Nov 16, 2026 | Which AI Use Cases Survive Contact with Real Workflows? |
| Nov 23, 2026 | Can You Reconstruct the Action? Evidence Requirements for Enterprise AI |
| Nov 30, 2026 | When AI Meets the Application Portfolio: Integrate, Replace, or Consolidate? |
| Dec 7, 2026 | Build, Buy, or Orchestrate? Enterprise Requirements for the AI Stack |
| Dec 14, 2026 | From Data Access to Data Use: What Must Change When Machines Become Consumers? |
| Dec 21, 2026 | Who Runs AI After Launch? Ownership, Evaluation, and Escalation |
| Dec 28, 2026 | Where Does Third-Party AI Extend Your Security Boundary? |
What we’re looking for
Across every conversation, we’re ultimately trying to identify three things.
What is breaking?
Where are existing enterprise architectures, economics, controls or operating models failing as AI moves into production?
Who cares enough to pay?
Which problems have moved from technical inconvenience to executive priority and funded initiative?
What needs to exist next?
What infrastructure, tooling or company doesn't exist yet — but increasingly needs to?
That intersection is where Mesa spends its time.
- Research
- Market Intelligence
- Companies
- Capital
What Breaks Next gives Mesa a systematic way to listen to the enterprise market. The research identifies emerging constraints. Those constraints reveal opportunities.
Mesa works with technically strong companies addressing those opportunities — helping them understand enterprise demand, sharpen positioning and build credible go-to-market strategies. And where we develop conviction, those relationships can inform how we think about investment.
That’s the purpose of What Breaks Next: find the problem before the market has a name for it.
Get in touch
What are you seeing break?
If you’re an enterprise technology leader, founder or investor working through one of these questions, we’d like to hear from you. Enterprise leaders can join the research; to discuss a Mesa engagement, tell us about it below.