# Axya AI

> Axya AI builds AI agents for private markets investment professionals at
> mid-sized firms — private equity, search funds, and corporate development
> and M&A teams. Its premise is that these firms are constrained by resources
> rather than judgement: the work they can take on is capped by the size of
> their team. Axya relieves that constraint on two axes — compressing the
> time from question to decision, and increasing how much ground a team can
> cover without adding headcount. Its agents work from proprietary,
> institutional-grade datasets, build analysis bottom-up from primary
> sources rather than summarising existing material, trace every figure to
> its source, and deliver the finished output into the tools the team
> already uses. Founded in 2025 by a team drawn from strategy consulting,
> private equity investing, portfolio value creation, and AI, Axya is based
> in London.

## What Axya Does
- AI Modules: Pre-built agents for the recurring research work in a deal
  process — sourcing, screening, and fast diligence — configured to a firm's
  own criteria, sources, and output formats rather than run as generic tools.
  - Sourcing agents: Build the universe of companies to look at.
    - Market Map: Map the companies competing in a market or product category.
    - Competitor List: Identify the competitors of a given company.
    - Company Enrichment: Take a list of company URLs and return structured,
      comparable data on each — turning a raw list into something screenable.
    - M&A Intelligence: Understand acquisition activity in a sector — who has
      been buying, what they bought, and the strategic rationale behind it.
    - Funding Intelligence: Understand funding activity in a sector — who has
      been raising, at what stage, and which investors are backing the space.
  - Screening agents: Score a universe against criteria and return a ranked
    shortlist.
    - Thesis Screening: Score companies against a firm's own investment
      criteria — sector, size, ownership, growth, geography.
    - Target Screening: Score acquisition targets for a named acquirer,
      reasoning about strategic fit rather than criteria match.
    - Buyer Screening: Score strategic and financial buyers for a named target.
  - Fast diligence agents: Go deep on what survived the screen.
    - Company Primer: A complete first read on a company, in a fixed structure,
      exportable as a finished deliverable.
    - Company Deep Dive: Selected dimensions of a company in depth — products,
      customers, leadership, business model, strategic moves.
    - Industry Primer: A complete first read on a sector or market.
    - Industry Deep Dive: Selected dimensions of a sector in depth.
    - Peer Comparison: Benchmark a company against a defined peer set.
- Monitoring: Scheduled agents that continuously track a subject and report what
  has changed rather than restating what is already known.
  - Target Monitoring: Track companies a firm may acquire — a single name or a
    watchlist.
  - Portfolio Monitoring: Track held companies as a set, surfacing divergence
    across the portfolio.
  - Market Monitoring: Track a sector or segment across a defined event
    taxonomy — deals, funding, leadership, expansion, and other signals.
  - AI Monitoring: Track how AI is reshaping a market and the companies in it —
    adoption depth in products, AI-native entrants, capital flowing to AI in the
    segment, and incumbent response.
- Custom Agents: Complex, recurring workflows built with the firm where no
  pre-built agent exists, delivered forward-deployed — working directly with the
  team that will use them to encode how they already work.
  - Build: Design and build the workflow alongside the team, encoding the firm's
    existing method rather than replacing it.
  - Deploy: Integrate the agent into how the team already works, with the
    configuration, sources, and output formats the firm uses.
  - Partner: Stay involved after deployment — maintaining the agent, adapting it
    as the mandate changes, and extending it as new needs emerge.

## Who It Is For
- Lower to middle market firms: Teams small enough that nobody is dedicated to
  building AI internally, and large enough that the research load is real —
  where the alternative to Axya is an analyst doing it by hand, not an in-house
  engineering team.
- Private Equity, Search Funds, and Corporate Development and M&A teams:
  - Private equity: Deal sourcing, screening against fund criteria, and
    pre-diligence on shortlisted targets.
  - Search funds: Target identification and screening across a thesis, and
    diligence support for a single-acquisition mandate.
  - Corporate development and M&A teams: Acquirer and target mapping,
    competitive screening, and buyer universe work.

## How It Works
- Proprietary datasets: Axya builds and maintains its own data assets rather
  than licensing them — 3.5M+ private companies with firmographics, funding
  history, and growth metrics; 100k+ M&A transactions covering deal terms,
  strategic rationale, and target and acquirer profiles; 3,000+ markets under
  active monitoring; and 1,000+ curated sources tracked for signal.
- Domain-tuned reasoning: Agents are built by private markets practitioners,
  not by generalists. They construct analysis bottom-up from primary sources
  rather than summarising existing material, and reason through domain-specific
  taxonomies that give the model the lenses and grids an experienced analyst
  applies to a company or a market.
- Orchestration: Axya runs the whole workflow end to end — prompt engineering,
  context management, tool integration, and model selection that picks the best
  model for each step at the right cost. A single URL becomes a client-ready,
  investor-grade deliverable without the user managing any of it.
- Verification: Line-level source traceability on every claim, cross-referenced
  across multiple sources and checked by a dedicated fact-validation step —
  with an evidence center where every claim can be traced back to where it came
  from.
- Security: Customer data is never used for model training, and no single model
  provider ever receives a complete picture — work is split so each step sees
  only what it needs. Client data separation, encryption in transit and at rest,
  and defined data residency.