Example package
AI proof of concept
One use case built on a sample of your real data, with the test results.
You get
- Use case agreed in writing
- Working build on your data
- Test set and pass rate
- Running-cost estimate
How it works
- 1You describe it
- 2A written scope
- 3Milestones you review
- 4Handover and a 30-day fix window
How we work, step by step · The 30-day fix window in our agreement (section 9)
Sound familiar?
What we hear from businesses before a project like this starts.
- Every AI development company we talked to wanted a six-figure discovery phase.
- We have a clear idea for an AI tool but no one in-house to build it.
- Our team pastes the same documents into ChatGPT every day.
- We tried an AI tool that worked in the demo and failed on our own data.
- We need to know what it will cost to run each month before we commit.
- Customer data can't leave the systems we already trust.
What buyers usually need
The requests that come up most often for this kind of work, and what each one takes to work properly.
01
An AI app for one job
A small web app that does one task well: read a document, draft a reply, check an order, summarise a call.
02
A custom AI agent
An agent that takes steps in your tools, such as looking up a record, updating the CRM or creating a task, with limits you agree in writing.
03
A chatbot on your own content
Answers from your documents and policies, with a hand-off to a person when it is unsure.
04
AI inside the tools you use
Models connected to your CRM, helpdesk, inbox or spreadsheets, so results land where your team already works.
05
A test set before launch
Your real cases with the expected result for each, run before launch and after every change.
06
A running-cost estimate
Model and hosting usage estimated from your volumes in the scope, billed to your own accounts.
Tools that come up most: OpenAI, Claude, Gemini, open models through Ollama, vector search, Python, Node.js, n8n, HubSpot, Salesforce, Google Workspace, Microsoft 365, Slack, Microsoft Teams, Zendesk.
What you get
Concrete pieces, each one something you can open and check when it's done.
The AI app or agent
Built around one job, with instructions, tools and limits written with you.
OpenAI, Claude, Gemini, Python or Node.jsYour data connected
The documents, records and systems it needs, connected with access you can remove.
Your CRM, drive, inbox or databaseA test set
Real cases and expected results, with a pass rate you see before launch.
Test sheet, run logGuardrails and an off switch
What it may and may not do, when it hands over to a person, and a way to pause it.
Admin settingsHandover
Source code in your repository, a run book and a walkthrough for your team.
Your GitHub or GitLabTypical projects
Example packages to show the shape of the work. These are examples, not past clients; yours gets its own written scope.
Example package
AI app or agent
A tool your team uses every day, connected to your systems.
You get
- Built in your accounts
- Guardrails and hand-off
- Test set run before launch
- Code and run book
Example package
AI chatbot
A chatbot on your website or in Slack or Teams that answers from your content.
You get
- Your content indexed
- Hand-off to a person
- Conversation log
- Weekly unanswered list
How we approach it
The habits that keep this kind of project from going wrong.
1
Small first
We start with one job that pays for itself, then build on it.
2
Tested on your cases
Results are checked against your real examples, not a demo script.
3
Your accounts, your code
Models, data and code sit in accounts you own.
How we work
A written scope first, then working progress you can check, then a proper handover.
You describe it
Tell us what you need in your own words, and which tools you already use. A few lines is plenty to start.
A written scope
We reply with a short written scope: what gets built, what is out of scope, which accounts we need, and how we'll both know it works. You know what you're getting before work begins.
Milestones you review
The work is split into milestones. At each one you see it working in your own account and sign it off before we go on.
Handover and a 30-day fix window
You get the working system, its documentation and a walkthrough. If anything we built breaks in the next 30 days, we fix it at no charge.
What shapes the scope
- How many systems it connects to
- How much data it works from
- Where it runs
- Expected monthly volume
Always included
- A written scope before work starts
- Milestones you review and sign off
- Documentation and a walkthrough at handover
- A 30-day window for fixes, at no charge
Pushbuild has built software for businesses since 2010, for more than 40 clients. Every project is led by a former public-company CTO, based in Massachusetts. See example builds.
Questions people ask first
What does an AI project with you look like?
A short brief, a written scope with milestones, then a first working version on your real data. You review each milestone before the next one starts.
Can you build a custom AI agent?
Yes. An agent gets a list of tools it may use, such as reading a record or creating a task, and limits we agree in writing. It is tested on your real cases before it acts on live data.
Where does our data go?
The scope names every system and model provider your data passes through. Where data can't leave your own systems, we use an open model on your own server.
What will it cost to run?
Model usage is billed to your own account. We estimate the monthly cost from your volumes in the scope.
What happens after handover?
For 30 days we repair anything we built that does not work as the scope describes, at no charge. After that, further work is scoped the same way and agreed with you before we start.
Do it once, we automate it. Show us the job one time and it runs on its own after that, in the tools you already use. Not sure where to start? Start with an Automation Audit.
Tell us what you need built.
Email a few lines about what you want and which tools you use, or call. We reply with questions or a written scope.
hello@pushbuild.com