AI technology consulting
We turn hard operational problems into working Gen AI systems
Senior-led AI consulting for operating businesses, with deep roots in energy. We build what we recommend.
The problem
of organisations in MIT's study saw no measurable return from Gen AI
The report points to weak integration with real workflows, not weak models, as the main cause.
Source: MIT NANDA, The GenAI Divide: State of AI in Business 2025 (preliminary report)
Our approach
Start from the workflow, not the model
-
01
One workflow
We pick a single workflow with a named owner and a measurable outcome, then choose the model.
-
02
Built by the people who advise
The people who advise on the design also write and ship the code, so the advice is tested in practice.
-
03
Proven on your data
Every pilot is scored on your own data and ends with a production plan.
Where we work
Problems we solve with Gen AI
Each starts from data you already hold. They apply across industries; our deepest experience is in energy.
Inspection and monitoring
Systems that spot defects, leaks and changes in photos, video and sensor images, ranked by urgency.
Knowledge and search
Assistants that answer from your own documents and show where each answer came from.
Reporting and workflows
Assistants that draft reports and handle routine checks, with a person approving the result.
Training and support
Bilingual tutors and training tools that answer only from approved company material.
Working agents we have built
Education
Admissions agent for schools
Answers parents in English and Arabic, books tours and assessments, and hands every lead to the admissions team.
Real estate
Lead agent for brokerages
Answers property enquiries on web chat and WhatsApp at any hour, quotes only permitted listings and books viewings.
Real estate
Floor plan to 3D
Reads a developer's 2D unit plan and turns it into a 3D model a buyer can walk through before a site visit.
For schools in the UAE and the GCC
An AI admissions agent for your school
Families research schools in the evening, in English and Arabic, and expect the fees, dates and documents exactly right. Our agent answers them day and night from your own documents and data, books tours and assessments, and passes every lead to your team.
Do you have Year 3 places for September 2027? My son was born on 14 November 2020.
AVA Children join a year group by their age on 31 August, so your son would start in Year 2, not Year 3. Al Barsha has one Year 2 place left for September 2027. Would you like to book a campus tour?
ELIGIBILITYYear 2 · age 6 on 31 Aug 2027
What are the fees, and is there a sibling discount?
What the agent does
Answers from your documents
Policies, prospectus, handbooks
Searches your school's own files in English and Arabic and names the document each answer comes from.
Answers from your data
Seats, fees, slots, applications
Checks live places by year group and campus, fees in AED, open tour slots and an application's status.
Takes action
Books, records, hands over
Books a tour, assessment or callback, saves the enquiry, and passes the conversation to a person when needed.
Built for how schools here work
| What parents need | How the agent handles it |
|---|---|
| Year-group placement | Age on 31 August, as KHDA and ADEK require, for British, American and Indian curricula |
| Arabic | A parent who writes in Arabic gets an Arabic reply, right to left, from Arabic documents |
| Fees in AED | Termly instalments, sibling and early-payment discounts, bus zones |
| Required documents | Emirates ID, attested transfer certificate and the rest of your checklist |
| Tour calendar | A Monday to Friday week with shorter Fridays, per campus |
Your team sees every lead and every step
Leads
Each family with child, year group and a short note written by the agent.
Bookings
Tours, assessments and callbacks, each with its reference number.
Handoffs
Questions only a person should answer, such as refunds, queued for staff.
Conversations
Every transcript beside a trace of what the agent looked up and did.
Knowledge base
Upload a PDF, Word, Excel or image file and the agent answers from it at once.
Usage and cost cap
A daily AI budget with a live meter, so running costs never surprise you.
Safe by design
The agent reads and writes only through fixed, checked actions. It never runs its own database queries and never invents contact details.
How a project runs
From a hard problem to working software
Short loops. You see working software early and can steer before much is spent.
- 01
Frame
Agree the problem, the data and what a good result looks like.
- 02
Prototype
Build a working version quickly with modern AI tools and show it early.
- 03
Prove
Score the results on your real data and fix what falls short.
- 04
Hand over
Add monitoring and documentation, and train your team to run it.
Reliability
How we keep AI output dependable
Gen AI can be confidently wrong. These four habits make results something you can rely on.
Grounded answers
Assistants answer only from approved sources and show them.
Evaluation suites
Every model or prompt change is scored on a test set before release.
Human review
People approve consequential decisions, and the system flags what it is unsure about.
Data handling
Your data stays within agreed boundaries, with access limited to the project.
How we engage
From pilot to production
The pilot is the point of commitment: working results before any large investment.
Phase 1
Discover
Choose 2 to 3 use cases and review the data you already hold.
Phase 2
Pilot
A 4 to 6 week proof of concept on your existing data, ending in a production plan.
Phase 3
Production
Integrate and extend what worked, once the pilot has earned it.
Who we are
Senior engineers, hands-on
AgentKrafts is a small, senior team of software engineers with deep AI expertise. The people you talk to are the people who build your system: no hand-off to a junior bench.
We have trained deep learning models, built assistants that answer from an organisation's own documents and data, and delivered AI use-case discovery and proofs of concept for large companies across the Middle East, with particular depth in oil and gas.
Next step
Bring us one hard problem
Tell us about one workflow or one data set. We will show working results in a short pilot, before any large investment.