NewAn AI cost-saving add-on for Claude and ChatGPT

Orchid Conductor™

The control room for your firm's AI.

The control panel between your people and every model. It protects, checks, routes and records each request, then shows and improves what AI delivers.

One controlled journey

Every AI request

Where your people work

Claude
ChatGPT
Copilot
Your own tools

Orchid Conductor™

What your firm gets

Sensitive details protected
Firm rules applied
A suitable, lower-cost model
A complete audit record
A fix ready for approval

Your database: conversations, content and records.

Orchid central: logins, usage, routing choices and counts.

01 / Control

Control every request.

One place to decide what can be sent, what is allowed, which model answers and what is recorded.

01

Protect confidential details

Find personal and confidential details, then mask, redact or block them before a model provider sees them.

02

Apply your firm's rules

Monitor or block requests by name and topic. Set access and budgets by team, role and person, with a full audit log.

03

Route for value

Send each task to the cheapest model that is suitable for it. Move automatically to another suitable model if a provider goes down.

04

Verify the figures

Check figures in an answer against their sources. Show what is verified, derived, unsupported or wrong, with a link back to the source.

02 / Understand

See what AI actually delivers.

Go beyond the AI bill. Understand the work, the time saved, the cost and where people still need a better answer.

01

Measure the return

See hours saved and their value, AI cost against the cost without Orchid, details kept private and adoption by team and person.

02

Find opportunities

Group repeated corrections and abandoned answers by role and kind of work, ranked by the hours at stake.

03

Explain what needs fixing

Distinguish a wrong method, missing data or a failing integration, so the next fix addresses the real cause.

04

Report to leadership

Create board-ready reports from the firm's own data, with findings, actions, owners and a record of whether each action worked.

03 / Improve

Turn every gap into better AI.

Build the fix, test it on real work and release it only after the right people approve it.

01

Create AI employees

Build reusable ways of doing named work, with your instructions, sources, checks and sign-off built in.

02

Connect your systems

Connect your CRM, deal database, underwriting engine, Microsoft 365 and public data. Every call carries the person's identity.

03

Pilot, measure, release

Start with a draft, test with a pilot group and compare against the same work without it. Senior approval comes before rollout.

04

Build models of your own

With enough examples and your explicit agreement, Orchid can train a model for your firm alone. Orchid deploys it, then it enters the same pilot process.

Your environment

Your content. Your database.

Every firm gets its own full install. Choose a dedicated private cloud, your own cloud or your data centre.

What stays with your firm

  • Conversations, prompts, answers and documents
  • Task records, rules and their matches
  • AI employees and their instructions
  • Reports, documents and files

What Orchid central receives

  • Logins, keys and your org chart
  • Usage and billing information
  • Routing choices and impact counts
  • Health reports, never your prompts or answers

Private cloud

The default dedicated install.

Your cloud

Your AWS or Azure account.

On-premises

Your own data centre.

Two ways to start

Start with oversight. Move to full protection.

Monitor mode

Everything Orchid can do without sitting in the middle of the request. People keep using Claude and ChatGPT as they do today.

  • Claude or ChatGPT answers, on the plans you already pay for
  • Every request and answer checked against your firm's rules
  • Your firm's style and terminology applied to every answer
  • Rules, access and teams set once from the org chart, inherited by every team
  • A searchable record of every conversation, kept in your own database
  • Your own systems connected, called with every call logged
  • Hours saved, adoption and what falls short, measured on every task
  • Board reports with owned actions, and AI employees your people can use
  • See whose AI use is being recorded, and whose isn't, team by team

Full mode

Everything in monitor mode, with every request passing through Orchid.

  • Sensitive details masked or redacted before any model sees them
  • Blocks enforced, never just requested
  • Each task routed to the best-value model, with automatic failover and lower cost
  • Figures in answers checked against their sources
  • Models trained on your own work, served for your firm alone

The complete control panel and the full product.

Common questions

The direct answers.

Do our people have to change tools?

No. They can keep Claude, ChatGPT or Copilot. Developers point their tools at the firm's Orchid address, and Orchid Studio is available for teams that want a full AI workspace.

Does Orchid see our prompts and documents?

Orchid's central service receives logins, usage, billing, routing choices and counts, not prompts, answers or documents. Conversations and content stay in the firm's own database.

Can we keep our Anthropic or OpenAI licence?

Yes. Bring your own model keys and keep your existing licence and terms. Conductor adds masking, your rules, routing and measurement on top.

Is masking perfect?

No automated detection is perfect. Conductor uses patterns, the firm's own fingerprinted list and a name-finding model inside the firm's install. You can block anything you cannot risk.

Orchid's network

Built with the right people around us.

Backed by

Investors in Orchid AI.

Ascension VC
Investor
CIBRA Capital
Investor
Black Seed
Investor
+4
Other investors

Supported by

Startup programmes and academic support.

Google for Startups
AWS Activate
Microsoft for Startups
LSE

Key partners

Commercial partners, separate from our investors.

See Conductor on your firm's real work.