# Custom AI agents to handle real work

Implementation service from Modern Projects. Custom AI systems for your firm's repeated work. We scope one workflow, build the agents, automations and integrations it needs on the software your team already uses, and run and improve them each month.

## What every build includes

### Built on your systems

What we build works in the email, files, CRM and models your team already uses. Your data stays where it is.

### Scoped permissions

Each system sees only what its users can see, and acts only with the access you grant it.

### Human review

You decide which steps run on their own and which wait for a person to approve.

### A record of every run

Each step is logged with its sources, so your team can check what was done and why.

## Our Agents

### Deal Intake Agent

Investments. Proactively processes every new opportunity, underwrites it using the firm's template and assumptions, and passes it off to an employee.

### Site Sourcing Agent

Development. Watches for zoning changes that fit the firm's criteria, tracks down the parcel owner and their contact details, and sends the first outreach.

### Buyers Log Agent

Investment Banking. Keeps a sale process moving: updates the buyers log as NDAs come in, provisions data room access, and drafts counterparty emails for the team to approve.

## Common questions

### Which workflows suit an agent?

Work that repeats often, starts from a clear trigger and ends in a clear output. Examples are reading deal packages into your pipeline, extracting rent rolls and operating statements, and tracking loan covenants. Work that depends on judgment we cannot check stays with your team.

### Which systems do you build on?

The systems your team already uses, such as Microsoft 365, Box, Google Drive, DealCloud, Yardi, CoStar and Altus Group. We connect through their APIs. Your systems stay the system of record.

### How do you build an agent?

We map the workflow with the people who do it now and agree what a correct result looks like. We build the agent in your accounts and test it on your past cases. Then we run a pilot on live work, with a reviewer checking each output.

### How do you check the agent's work?

Before launch, we test the agent against a set of your past cases. Each output links to its source documents. Outputs that fail a check go to a reviewer. A step runs without review only after it meets the agreed accuracy on live work.

### What can an agent access?

Only the data and actions you approve, within the permissions of the people it works for. Actions with an outside or financial effect wait for a person to approve them. Every run is logged.

### Who owns what you build?

You do. It runs in your accounts, and we hand over the code, configuration and documentation. Your team or another provider can run it without us.

### How do you scope the work?

We scope one workflow at a time. Each scope states the current baseline, the acceptance criteria, what is excluded and the timeline. We fix issues for 30 days after launch.

## Contact
- [Book an introduction call](https://cal.com/jackson-davis-7ojpba/modern-projects-introduction-call)
- Email: [founders@modernprojects.ai](mailto:founders@modernprojects.ai)
- Site guide for AI agents: [llms.txt](https://www.modernprojects.ai/llms.txt)