Thesara AI / Opportunity
The enterprise software layer is being re-tendered for the fourth time in fifty years. Thesara AI is building for that transition on two fronts — its own industry suites, and the services that carry other organisations across. This page sets out the market, the model, the plan and the risks.
The market
The figures below are published third-party estimates, attributed to their source. They describe the size of the transition. Thesara AI's claim is not that the transition is coming — everyone can see that — but that most organisations cannot cross it without help.
Gartner estimates that up to $234 billion of enterprise application spending is exposed to agentic disruption through 2030 — roughly 20% of enterprise application SaaS spend by that year — as agents complete work across systems rather than people clicking through them.
Source: Gartner, July 2026
Gartner projects agentic AI software spending reaching approximately $985 billion by 2030, a compound annual growth rate above 60% from 2025 — and suggests agentic AI could account for close to 30% of enterprise application software revenue by 2035.
Source: Gartner, 2026
Gartner expects around 40% of enterprise applications to embed task-specific agents by the end of 2026, up from under 5% in 2025. Buyers are committing well ahead of their own ability to deliver — which is precisely where a services practice earns its place.
Source: Gartner, 2026
Gartner expects more than 40% of agentic AI projects to be cancelled by 2027. We treat this as the central fact of the market rather than an inconvenient one: the opportunity is not enthusiasm, it is execution.
Source: Gartner, 2026
Revenue model
Services revenue arrives early, is cash-generative and requires little capital — which pays for the product engineering that produces higher-margin recurring revenue later. Holding both smooths the cash curve that sinks most software ventures.
| Line | Revenue source | Contract shape | Margin character |
|---|---|---|---|
| Services | Readiness assessments | Fixed fee, 3–5 weeks | Low by design; opens the account |
| Services | Custom AI development | Fixed scope per phase, milestone-billed | Moderate; predictable and staffable |
| Services | Legacy migration | Phase-gated fixed price, 4–12 months | Moderate; largest single engagements |
| Services | Managed operation | Monthly retainer | High and recurring; grows without headcount |
| Products | Suite subscriptions | Annual per seat, site or outcome | High, recurring, expands within accounts |
| Products | Platform & marketplace | Usage-based, plus partner revenue share | High, and scales independently of delivery capacity |
Why it defends itself
A client whose workflows, permissions, audit evidence and exception queues run on our platform is not migrating on a whim — the cost is not the licence, it is retraining the organisation.
Where the moat is
Anyone can call a model. Very few can encode how a pharmacy adjudicates a claim, how a mill costs a sample, or what an examiner will accept as an explanation — and that knowledge accumulates with every engagement.
What we refuse to be
A thin interface over someone else's model is a feature, not a company. The defensible parts are the trust layer, the evaluation harness and the industry logic — which is where our engineering goes.
Capital plan
This is a software business, so most of the capital is people. Capital is released against phase gates — a shipped suite, a signed reference customer, a met revenue test — rather than against a calendar.
Figures below are the allocation framework — confirmed amounts are in the information package
| Application | Indicative allocation | What it buys |
|---|---|---|
| Platform & product engineering | 40–50% | The shared core and the first waves of industry suites through to general availability |
| Services delivery capability | 15–20% | Delivery engineers, solution architects and industry specialists |
| Compute & model costs | 8–12% | Cloud and specialist compute, model access, training and evaluation runs |
| Certification, legal & compliance | 6–10% | SOC 2 and ISO programmes, regional entities, regulatory counsel, insurance |
| Commercial & go-to-market | 10–15% | Industry sales, partnerships, customer success and onboarding |
| Working capital & contingency | 8–12% | Operating runway, hiring buffer, cost inflation |
| Total | 100% | Released in tranches against phase gates |
Participation
Participation is through Thesara AI's designated private member association, under a relationship and revenue-sharing covenant. Every step below happens before any capital moves.
You submit an LOI through this website. It is a request for information and a statement of interest — not a commitment, not binding, and no funds are involved at this stage.
A named member of the Thesara AI team contacts you, confirms eligibility for your jurisdiction, and answers questions before anything is sent. Identity and source-of-funds checks are completed here.
Full disclosure set: association documents, covenant terms, financial model, product roadmap, risk register, and the reporting schedule you will receive as a member. Reviewed with your own advisers, at your own pace.
If you proceed, membership and covenant documents are executed. These documents — not this website — govern the relationship entirely, and they control over anything published here.
Distributions are made on the covenant schedule. Members receive quarterly operating reports, annual audited financial statements and access to the member portal — with the same figures available to any auditor a member appoints.
Risk
A prospectus that lists only upside is not a prospectus. These are the risks we consider material, the honest version of each, and what we do about it. The full register is in the information package.
The risk: Gartner expects more than 40% of agentic AI projects to be cancelled by 2027. If enterprise buyers retrench, sales cycles lengthen, pilots stop converting, and both revenue lines slow at the same time.
Mitigation: every engagement is sold through a measured pilot with an agreed success test; pricing anchored to outcomes the customer can verify; seven suites across unrelated industries so no single vertical's slowdown is fatal.
The risk: the companies supplying the models could build industry applications themselves, or general-purpose agents could become capable enough that specialised software matters less.
Mitigation: our defensible layer is not the model. It is the industry logic, the trust and audit layer, the evaluation harness and the integration into systems of record — none of which a general model supplies. We remain deliberately model-independent so that improvements upstream benefit us rather than displace us.
The risk: we buy compute and model access from third parties. Price increases, capacity constraints, licence changes or a provider deprecating a model we depend on all affect our margin directly.
Mitigation: model routing that selects the cheapest model meeting the quality bar per task; the ability to run open-weight models we host ourselves; multi-provider contracts rather than a single dependency. This is also the risk that, at sufficient scale, argues for owning infrastructure later.
The risk: services revenue is easier to win than product revenue, and a company can drift into being a consultancy — growing headcount linearly with revenue, owning nothing reusable, and never shipping the products.
Mitigation: a standing rule that fixes go into the shared platform rather than a client fork; a board-level target for the ratio of recurring to project revenue; reported to members quarterly so drift is visible early rather than discovered late.
The risk: incumbent application vendors are embedding agents into products that already hold the customer's data, and the global systems integrators have far greater delivery capacity than we will for years.
Mitigation: we compete on depth in seven chosen industries rather than breadth; on shipping AI-native architecture rather than retrofitting; and on being small enough to serve buyers the large integrators will not prioritise. Partnership with both categories is expected, not excluded.
The risk: AI regulation is moving quickly and unevenly, particularly for employment, clinical and credit decisions — exactly the areas our suites operate in. Rules may raise costs, restrict features, or expose us to liability for decisions our software influenced.
Mitigation: human approval required wherever a decision affects health, employment, education or money; decision logging built into Layer 1 so new audit requirements do not require re-engineering; local counsel per region; professional indemnity insurance sized to the exposure.
The risk: we process clinical, financial and employment data on behalf of clients. A breach would be materially damaging to revenue, reputation and standing in regulated markets.
Mitigation: tenant isolation, encryption, customer-managed keys, documented residency per workload, SOC 2 and ISO programmes, and a contractual commitment that client data is never used to train models for anyone else.
The risk: this is a talent business in the most competitive hiring market in technology. Losing a small number of senior engineers or industry specialists would slow delivery materially, and early revenue may be concentrated in a few clients.
Mitigation: documented architecture and delivery method rather than knowledge held in heads; long-term incentives tied to the covenant; explicit customer-concentration reporting to members each quarter.
The risk: member participation is not publicly traded, transfer is restricted, distributions depend on operating performance, and multi-currency operations carry exchange exposure.
Mitigation: transfer and redemption terms are stated in the definitive documents and should be read before commitment; participation is intended only for those who can hold an illiquid position for its full term; currency policy is set by the board and reported.
Begin
The Letter of Intent & Request opens the private member channel and releases the full information package for your review.