Agents that finish the job, with a person in the loop.
An agent gathers context, uses tools, takes a bounded action and hands work back when judgment matters. We add approvals, logs and clear failure paths so it is safe to run in production.
Custom AI software that cuts tool spend and busy work.
AI agents and internal tools, built around the work you already do. Fixed scope, fixed price, no per-seat fees.
- Free audit first
- You own the software
- Training included
Six products we designed, built and shipped.
No borrowed client logos. These are our own products, and most of the code is public.
See the work →- Karko AI
- Nura RAG Copilot
- Sadaa
- Bunyan
- Kursfind AI
- Swipekit
What you can hire us for.
Four services. One accountable team. A fixed-scope build when the workflow is clear, or team training if you want to start smaller.
- 01
Custom AI agents
Software that does a real job end to end, with guardrails and a human in the loop when judgment matters.
- 02
RAG (answers from your company documents)
Your documents and data, answerable in plain language, with citations and a clear “I don’t know” when evidence is missing.
- 03
Custom internal tools
Dashboards, admin panels and ops tools your team uses daily. AI-powered where it helps, plain software where it doesn’t. You own the source.
- 04
AI training for your team entry point
Get your team productive with Claude Code, Claude Cowork, Cursor, ChatGPT and Codex, plus training on whatever we build. The lower-commitment place to start.
Book a free audit
Free audit first. We’ll recommend a build or a training start.
Busy work has a price. See yours.
Pick the jobs that eat your team's week. We'll show what one owned build takes back each year, in hours and dollars.
the busy work
people
- Hours a build takes back, per year
- ~1,877 hrs
- Worth, at $40 an hour
- $75,072
one owned build
we'd build: support agent · data-entry agent · document answers
- Build fee, fixed scope
- $30,000
- Running cost, per year
- $7,680
Hosting, AI usage and upkeep. Not per seat.
You keep, per year $67,392
Book a free audit We'll use your real numbers in the audit.
How we calculate this
- Each job uses a preset far below survey benchmarks. Example: people doing heavy data entry average 9 hours a week on it (Parseur, 2025); we count 4 hours, for a quarter of the team, and only 60% of that as removed.
- Around 60% of workers say automation would save them 6+ hours a week (Smartsheet). Across a whole company we count roughly 1 to 2 hours per person.
- Time is priced at $40 an hour fully loaded, below the US average employer cost of $46.60 (BLS, 2026), for 46 weeks a year.
- Build fee: $18k for the first job, $6k per extra job. In line with 2025-2026 market data for custom AI builds at this scale ($15k to $60k).
- Running cost: $150 a month for hosting and AI usage, $8 per person a month, plus 10% of the build fee a year for upkeep. Not per seat.
- If the math says don't build, we'll say that too.
Live updates need JavaScript. The numbers above show our worked default: a 30-person company with three jobs selected.
Built around the job, not the model.
How the three kinds of builds work in practice. Training is separate: it's for your team, not a system we ship.
Answers grounded in your documents and data.
Your sources become searchable in plain language. Every answer cites evidence. When the evidence is not there, the system says so instead of guessing.
Ops software your team will use every day.
Dashboards, admin panels and internal apps built around the people who run the work. AI where it earns its place. Plain and fast everywhere else.
Every build includes training for the people who will run it.
Proof before promises.
We’re a new studio, so these are not client case studies. They are products our team designed, built and shipped. Most are open source, so you can inspect the work yourself.
product / built by Useful Build
Visit Karko AI ↗ Karko AI
An AI agent that runs a whole job search: finds roles, checks fit and writes the application from your real CV. A full application takes about 90 seconds.
product / built by Useful Build
View the repo ↗ Nura RAG Copilot
A support copilot that answers only from retrieved documents and cites every source. When the evidence isn’t there, it says so.
product / built by Useful Build
View the repo ↗ Sadaa
Local-first voice dictation for macOS. Tap a key, speak English or German, and the words land at your cursor.
product / built by Useful Build
View the repo ↗ Bunyan
A macOS terminal for running many projects and AI coding sessions in one window. It runs our own work every day.
product / built by Useful Build
View the repo ↗ Kursfind AI
AI course discovery for the German education market, with recommendations, applications, provider dashboards and bilingual support.
product / built by Useful Build
View the repo ↗ Swipekit
Turns a plain script into an on-brand LinkedIn or Instagram carousel with AI backgrounds and one-click export.
From stuck workflow to working system.
It starts with a free audit. One accountable team stays close from there to the handover.
01
Scope
The free audit maps the workflow, names what you already pay for, defines the useful first release and agrees cost and date before we write code.
02
Build
You see working software early. We test it against the real job, not a slide deck.
03
Train
Your team gets documentation and working sessions before we call the project done.
04
Support
After launch, an optional retainer covers monitoring, fixes and improvements. No lock-in.
Start with a free audit. If we build, a fixed project fee gets you to launch. The Savings Blueprint above and the audit give you real numbers on cost and savings.
Book a free auditTell us the workflow and the tools you rent today. We’ll tell you the sensible next step.
A small team, close to the work.
Useful Build is led by Wasim Jalali, founder and AI specialist. He works with a small team to scope, build, train and support each project. You stay close to the people doing the work.
Wasim Jalali Founder and AI specialist
one accountable team
Useful Build
- Scope
- Build
- Train
- Support
What we put in writing.
These are not marketing claims. They are terms we actually use.
01
Truthful about capability
If we can’t build it well, we’ll say so before you spend anything.
02
Clear scope up front
What you get, what it costs and when it lands are agreed before we start.
03
Your data stays yours
We don’t train on it, resell it or move it anywhere you haven’t agreed to.
04
Fair terms
Plain contracts, no lock-in and no surprise line items.
05
AI only when it helps
Sometimes a script beats an agent. We’ll recommend the script.
What buyers usually ask.
01 What does Useful Build build?
Useful Build designs and builds custom AI software for small and medium businesses. That includes AI agents, RAG systems that answer from company documents, internal tools and practical team enablement.
02 Who is Useful Build for?
Useful Build is for small and medium businesses with repetitive work, disconnected software or expensive tools that do not fit the job. The primary market is the United States and Canada.
03 What is the free audit?
The free audit is a 30-minute working session with no obligation to hire us. We review the workflow that hurts and the software you already pay for, then explain what is worth building, what to leave alone and where you may be able to cut spend.
04 What does a build cost?
Every build has a fixed project fee agreed before work starts. The amount depends on the workflow and the useful first release, and you get a specific number after the free audit.
05 Can this replace tools we already pay for?
Sometimes. The free audit maps your real software stack and shows which overlapping tools may be retired, which should stay and what a custom build would cost.
06 What does RAG mean?
RAG means retrieval-augmented generation. In practical terms, it lets an AI system find relevant information in your approved company documents and use that information to answer with better context.
07 What is an AI agent?
An AI agent is software that can carry out a defined workflow using tools, rules and company context. It can do more than chat, but it should still operate within clear permissions and review steps.
08 How long does it take?
The timeline depends on the workflow and integrations. We agree a delivery date before code starts and show working software early.
09 Who owns what you build?
You own it. You receive the source code, documentation and training needed to run it without us.
10 Where does our data go?
Your data goes only to systems and providers you approve for the project. We do not train on it, resell it or move it without your approval.