Start a project

AI built around your team’s work.

We build AI systems that help your team find answers, serve customers, and complete tasks. We connect them to your tools and data, with your team approving actions where needed.

Selected clients

What we build

Business process automation

We use AI to review information, prepare work, and complete tasks across your tools, with your team approving actions where needed.

For example

  • Prepare a report from current records and send it for review
  • Update task records as work moves through an agreed process
  • Draft follow-up messages for your team to approve

How we build it

We map the process with your team and define each step. We connect the tools it needs and add approval checks for agreed actions. We record completed steps and send exceptions to a person for review.

Where we’ve built it

Voice and chat assistants

We build voice and chat agents that help people ask questions and complete tasks, with a clear path to your team when they need help.

For example

  • Help callers check availability and request a booking
  • Answer customer questions in website or messaging chats
  • Collect the details your team needs and pass the conversation to a person

How we build it

We connect voice and chat to your approved information and tools. We test conversations, response times, and interruptions. We define when the assistant should ask for confirmation or pass a request to your team.

Where we’ve built it

Search and answers from your data

We build search and answers across your documents and systems, with sources you can check and clear limits when information is missing.

For example

  • Answer from policies, contracts, and manuals, citing the page and the version
  • Look up current records in your connected tools
  • Say when sources disagree, or when the answer isn’t there

How we build it

We connect approved documents and records to a search system. Access checks limit what each person can retrieve. We test whether answers use the right sources and flag missing information.

Where we’ve built it

AI connected to your systems

We design how models, tools, and data work together. We connect them to your existing systems, with checks between steps.

For example

  • Read current customer or project records from your existing tools
  • Update records after the required approval
  • Pass information between your AI assistant, calendar, and task system

How we build it

We connect AI to the tools and data each task needs. We define what it can read and change, then build access checks and approval rules into those connections. We test failed requests and retries before launch.

Where we’ve built it

Custom software development

We build custom products and internal tools, from interfaces and features to the backend and infrastructure, with AI where it helps.

For example

  • Build web products with accounts, payments, and AI features
  • Create internal tools for schedules, reports, and approvals
  • Add custom interfaces and features to an existing product

How we build it

We design and build the interfaces, application logic, and data storage. We connect existing systems where needed and test the main tasks before launch. We add AI where it helps the work.

Where we’ve built it

AI performance optimization

We test how your AI handles the work you expect it to do. We use the results to improve accuracy, consistency, response time, and cost.

For example

  • Test changes to prompts, models, and tools against real tasks
  • Compare accuracy, consistency, response time, and cost
  • Track recurring failures and check whether updates resolve them

How we build it

We build tests from real tasks and compare the results with your team’s expectations. We check proposed changes before release and monitor results over time. Your team can review the tradeoffs between quality, speed, and cost.

Where we’ve built it

Ways to work with us

Bring us an idea, a defined project, or a system you want to improve. We’ll agree on the scope, deliverables, and support before work starts.

  • Plan your project

    We review your process and test a key assumption with your data. You receive a project scope, tests to measure success, and a build plan you can use with us or another team.

  • Build your system

    We build your system in agreed stages. At each stage, you receive working software tested against the tasks we agreed with you.

  • Run and improve it

    We host and monitor your system, test its performance, update models, and make agreed improvements. This is a monthly service. Your proposal defines the included work and usage limits.

  • Train and advise your team

    We train your team to use AI coding tools, choose models, and test results. We also help improve your codebase, security, and the process for releasing software.

What we build with

We choose technology around your tasks, data, and existing systems. These are tools we have worked with. We agree on the right setup for each project.

  • AI models and agents

    We build assistants with TypeScript, Vercel AI SDK, Eve, and AI Gateway. Our projects use OpenAI and Google Gemini models. We test models on your tasks and compare quality, cost, and response time.

  • Web and mobile apps

    We build web products with React, Next.js, TypeScript, and Webflow. For mobile apps, we use Expo and React Native. Our native iOS prototype work uses Swift and SwiftUI.

  • Data and databases

    We work with PostgreSQL, SQLite, MySQL, Firebase, and MongoDB. We also connect, process, and search data from files, documents, and external systems.

  • Voice and messaging

    We use speech models and WebRTC for voice experiences. We connect assistants to WhatsApp, Telegram, and Slack. We define when a person should review or take over a conversation.

  • 3D experiences and maps

    We build browser walkthroughs with Three.js and React Three Fiber. We prepare models and assets with Blender and Python. We use Google Maps and Turf for maps and location features.

  • Cloud and background jobs

    We use Vercel, Google Cloud Run, and AWS to host systems, with Docker where needed. We run background tasks with Trigger.dev, Vercel Queues, and Vercel Workflow. We choose the setup to fit the work and agreed operating limits.

  • Accounts, payments, and email

    We use Better Auth for accounts and access, Polar and Stripe for payments, and Resend and Postmark for email. We connect these services to the product’s records and processes.

  • Testing and monitoring

    We check behavior with Vitest, the Node.js test runner, and TypeScript checks. We use PostHog and Vercel Analytics to understand product use. We test changes and monitor the results that matter to your team.

Common questions

Do we need to prepare our data first?

You can start with the documents, tools, and records you have. We identify missing information, access issues, and any preparation needed when we plan the project.

Which models do you use?

We choose models based on quality, cost, and response time on your tasks. Where practical, we build connections that let us change providers. We test changes before switching.

Who owns the finished work?

The custom code, prompts, test cases, and documentation are yours under the project agreement. Your data stays yours. Third-party tools retain their own licenses. We can run the system or hand it over.

How do you handle privacy and security?

We limit access to what each task needs. We agree on provider data use, retention, and hosting before connecting client data. Where required and supported, we use zero-retention options or deploy in your cloud.

Read our Data processing addendum and privacy policy.

Can our team approve actions before they happen?

Yes. We agree on which actions need approval and build those checks into the tools that send, pay, delete, or publish. Requests outside the agreed scope go to a person.

What if AI isn’t the right answer?

We’ll say so during discovery. A rule, a form, or a process change may solve the problem with less cost and maintenance.

Tell us what needs to work.

Describe the task, who does it, and what you want to improve.