Pegasus agents · Custom agents
Your operations team shouldn't be the glue between your systems.
We build AI agents that do that work. Pegasus agents are ready for privacy requests, the month-end close, contract review and onboarding. Custom agents are built for the workflows that are yours alone. Your people keep the decisions.
Building your own agents? See the developer stack
Pegasus at work · illustrative run, sample data
Prompt:
running
—Past delivery work for
The problem
You hired operators. They spend the week re-keying.
A growing company runs on systems that were never built to talk to each other. The gap between them is filled by people: copying a record from one screen to another, chasing an approval, assembling the same report every month.
The work has to be done, and it is skilled people doing it. It also grows with every customer, every hire and every new rule, so the only way to keep up has been to add headcount.
Privacy requests
Every request means searching each system for one person's data, redacting it and replying before the deadline.
Month-end close
Transactions matched by hand, approvals chased and reports assembled from systems that do not agree with each other.
Contract review
Every contract read line by line for the same risk clauses while the deal waits.
New-hire onboarding
One new hire needs steps from HR, IT, facilities and a manager, and someone has to chase each of them.
Why it is still a problem
You have probably tried to fix this already.
A workflow tool
It automates the steps you can draw as a flowchart. Someone still has to build and maintain every flow, and the judgment calls stay with your team.
A strategy deck
You get a roadmap and a list of use cases. Nothing is running at the end of it, and the work between your systems is unchanged.
A general AI assistant
It can draft and summarize. It does not work inside your systems, and you cannot show an auditor what it did or why.
What is missing is something that does the work itself, inside your systems, and leaves a record a person can check.
What changes
The agent does the work. Your team handles what needs judgment.
01
Packaged or custom
A Pegasus agent is already built for a common workflow, such as privacy requests or the month-end close. When the workflow is your own, we build a custom agent for it.
02
Inside the systems you already run
The agent connects to the platforms your team uses today. Nothing is replaced.
03
People keep the decisions
The agent does the repetitive steps and sends a person only the exceptions, the risks and the approvals.
04
Everything on the record
Every action is logged with the data it touched and what it decided, so you can answer the question an auditor will ask.
You start with a pilot: one workflow, on your own systems, before anything wider. You judge it on your own work.
Where teams start
Pick the workflow that costs your team the most hours.
Four Pegasus agents are ready to pilot today. If your workflow is not one of them, we build a custom agent for it.
Privacy, legal and compliance
Compliance Agent
Answers data subject access requests: intake, identity verification, finding the data across connected systems, redaction and the response.
See how it runs →
Finance
Finance Agent
Works the month-end close: transaction matching, expense categorization, approval routing and the close task list.
See how it runs →
Legal and procurement
Contract Agent
Reviews contracts against your playbook: extracts the key clauses, flags the deviations and writes the review summary.
See how it runs →
HR, IT and hiring managers
People Agent
Runs new-hire onboarding: document collection, system provisioning, benefits enrollment, training and manager notifications.
See how it runs →
A workflow of your own? We build custom agents. You can also see all ten solutions or assess your workflow.
How we build
The agent prepares. A person decides.
Talking News Federation
The Talking News Federation, a volunteer-run membership organization, runs its member support on an AI help desk we built. Each inbound email becomes a ticket, is classified and routed, and has a reply drafted before anyone opens it.
A person approves every reply. Every AI decision is written to a tamper-evident audit trail, and suspected spam is quarantined for review, not deleted.
For developers
Your agent works in the demo. Production is where it breaks.
If you are building your own agents, the model is rarely the hard part. The agent needs somewhere to run, limits on what it may do, a way to pick up after a crash, and a way to get cheaper to run.
That is four products that work as one stack. Two run on your machine today as local developer previews. Two are in development.
- RunIthacaIn development · waitlistVirtual machines for AI agents
- ControlRemusLocal developer previewPermissions and approvals for AI agents
- RecoverRomulusLocal developer previewCrash recovery and durable execution for AI agents
- ImproveDaidonIn development · early accessEvaluate and fine-tune smaller open models on your own data
Is this for you?
Built for the teams doing the work, and the teams building agents.
Wherever people are the link between systems, there is a workflow to start with.
Direct answer
New Odyssey builds AI agents for any team whose people spend their time moving work between systems, whether that is finance, compliance, legal, HR or operations. Pegasus agents are packaged for common workflows; custom agents are built for a workflow that is your own. Either way the agent works inside the systems the team already uses and sends a person only the exceptions and approvals. For engineering teams building their own agents, New Odyssey makes a four-product stack: Ithaca to run them, Remus to control them, Romulus to recover them and Daidon to improve the models behind them.
Insights
Enterprise AI without the hype.
Guides & playbooks
From the blog
FAQ
Frequently asked questions
AI agents for operations teams, built two ways. Pegasus agents are packaged for common workflows such as privacy requests, the month-end close, contract review and onboarding. Custom agents are built for a workflow that is your own. Ithaca, Remus, Romulus and Daidon are for developers and engineering teams building their own agents: Ithaca runs them, Remus controls what they may do, Romulus recovers them after a failure and Daidon evaluates and fine-tunes smaller open models on your own data.
Pegasus is available: some agents are pilot-ready, others are deployed together as suites, and some are in development and labeled as such. Remus and Romulus are local developer previews you can download and run on your own machine. Ithaca is in development with a waitlist. Daidon is in development with early access.
Yes, and a product company too. As an AI consultancy we find where AI belongs in your operations and build custom agents for your own workflows. We also sell Pegasus, agents already packaged for common workflows, which you can start with a pilot on one real example.
A vendor's agents work inside that vendor's platform. Pegasus agents work across the HRIS, ATS, accounting, CRM, ticketing and file storage systems you already run, connected through Merge's unified APIs with no custom integration build.
Every Pegasus agent runs under the same governance layer: every action is audited, every deployment is versioned and every decision is traceable. Each agent holds at defined approval points, and a person handles the exceptions.
That is the direction: run, control, recover, improve. Today they are built separately. Remus is not yet integrated with Ithaca, and a shared SDK, trace and billing experience are not available yet.
New Odyssey Ltd is registered in England and Wales and based in London, with US operations in Dallas, Texas.
Next step
Start with one workflow.
Pick the process that costs your team the most hours. We pilot a Pegasus agent on it, on your own systems, and you judge it on your own work.