AI strategy consulting
AI strategy consulting that ends in a working pilot
A ranked, costed AI plan your board can check, written by senior engineers. The same engineers can build the highest-ranked idea as a paid pilot, or you can take the plan to another team.
For CEOs, COOs and CTOs whose board wants an AI plan
- Free 30-minute call with the founder
- Paid pilot you can stop
- You own the plan and the code
What is AI strategy consulting?
By the HorizonLux engineering team · reviewed by Seif Sgayer, Founder · updated 30 Sep 2026
AI strategy consulting decides where a company should use AI and in what order: HorizonLux gives CEOs, COOs and CTOs a ranked, costed plan their board can check. Senior engineers score each AI idea in scope by hours saved, effort, data readiness, risk and monthly running cost, and say plainly which ones should stay manual.
The plan is one written document in four parts: readiness notes, a ranked shortlist, an AI roadmap and AI governance guardrails. If you go ahead, the engineers who wrote it build the highest-ranked idea as a paid pilot, with one agreed deliverable and a demo every week. If it doesn't land, you stop. Every document and all the code stay in your own repo.
HorizonLux is a software and AI engineering company, founded in 2020 and based in Tunisia, working with clients in the UK, UAE and EU. Its senior engineers work in English, in your working hours on request. Every engagement starts with a free 30-minute call with the founder, Seif Sgayer, where price and length are agreed.
HorizonLux at a glance
- Company
- HorizonLux, a software and AI engineering company
- Based in
- Tunisia
- Founded
- 2020
- Works with
- Clients in the UK, UAE and EU
- Delivered by
- Senior engineers only, in English, in any timezone
- How it starts
- A free 30-minute call with Seif Sgayer, Founder. Price and length are agreed there
- You own
- The code, the IP and the docs, all in your own repo. NDA on request
Three questions your board will ask before it backs an AI plan
Only 27% of executives have a comprehensive AI strategy, Gartner reported in May 2026. Published research names three common reasons AI projects stall, and each one is a fair question for a board to ask. A plan worth backing answers them with your own numbers, not industry averages.
- 84%
of industry interviewees in RAND's study named leadership-driven causes of AI project failure. The most common root cause: misunderstanding what problem AI should solve. Board question: are we solving the right problem?
Source: RAND, 2024
- 63%
of organisations lack, or are unsure they have, the right data practices for AI. Board question: is our data ready?
Source: Gartner, 2025
- About 1 in 5
respondents to McKinsey's 2026 survey say AI running costs, including tokens, already limit their AI use. Board question: what will it cost to run?
Source: McKinsey, 2026
Sources: Gartner, press release on people-centric AI strategy, 13 May 2026; RAND, "The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed", Aug 2024; Gartner, "Lack of AI-Ready Data Puts AI Projects at Risk", Feb 2025; McKinsey, "The state of AI in 2026: On the road to ROI", Aug 2026. Checked 30 Sep 2026.
What you get from AI strategy consulting
Four parts make up one written plan, not a slide deck. The pilot and the board update follow only if you go ahead.
Readiness notes
The written result of the AI readiness assessment: where your data lives, how usable it is, and which systems each idea would have to connect to.
- Data sources and owners
- Systems and integration points
- Gaps to close, each with an owner
A ranked shortlist
Every AI idea in scope, from staff, vendors or the board, scored the same way and put in order, with a monthly running-cost estimate for each. Ideas to park or keep manual are listed too, with the reason.
Pilot nowNextLaterKeep manual
A written AI roadmap
The order to build in, and what must be true before each next step starts.
- An owner on your side for each step
- What to measure at each step
- The scope of one pilot, with its deliverable
AI governance guardrails
Written rules for how each use case may run, set before anything is built. The NIST AI RMF and ISO/IEC 42001 serve as reference points, not a certification.
- Human review points
- Data rules per use case
- A starter AI use policy
A working pilot
The highest-ranked idea, built in your systems by the engineers who wrote the plan, with one agreed deliverable. Your team tests it on real work.
- Code in your repo
- Docs, runbooks and training at handover
A board update
A short written note for the board: what ran, what it showed, what it costs to run and what to fund next.
- Results from real work, not forecasts
- The next decision, stated plainly
Plain English throughout, so your leads and your board can read every page.
AI use case prioritization, shown on a sample plan
This is the ranked-shortlist page of a sample plan, with made-up numbers. Every score shows the assumption behind it, so your board can question it and your team can update it. An idea that saves many hours but needs months of data work moves down.
| Score | Support triage | Invoice matching | Demand forecast | Sales email drafts |
|---|---|---|---|---|
| Hours saved a week | 34 h. Assumes 2,200 tickets a month, 4 minutes saved on each | 21 h. Assumes 900 invoices a month, 6 minutes saved on each | 18 h. Assumes two planners each save 9 hours a week | 5 h. Assumes five reps each save an hour a week |
| Effort to build | Low. One helpdesk and one queue to connect | Medium. The accounting system and the invoice inbox to connect | High. Sales history is split across three systems | Low. Drafts appear inside the CRM |
| Data readiness | Ready. Two years of tagged tickets | Mostly ready. Some invoices lack an order number | Not ready. No single sales history yet | Ready. CRM notes and past emails |
| Risk, and its guardrail | Medium. Refund and complaint tickets always go to a person | Medium. No payment is made without a finance sign-off | High. Planners keep the final say on stock orders | Low. Reps review and send every email themselves |
| Running cost, US dollars a month | About 380. Two model calls per ticket, at list prices | About 120. One model call per invoice | Not estimated. Scored again once the data is fixed | About 240. One model call per draft |
| Decision | Pilot now. Most hours saved, data ready, low effort | Next. Starts once the pilot shows results | Fix the data. Join the sales history, then score again | Keep manual. Few hours saved, and reps say personal emails win replies |
Sample for illustration, not client data. In a real plan, the roadmap page then sets the order and an owner for each step, and the guardrails page turns each risk line into a written rule.
From AI readiness assessment to a pilot your team has tested
Nothing is built until you sign off the plan, and every step ends in something you keep.
- 01Free
Free 30-minute call
Tell Seif what the board asked for and which ideas are on the table. You decide together whether a plan is worth doing now, and agree its scope, price and length.
You get · An agreed scope, or a plain "not yet"
- 02
Readiness assessment
Our engineers review the data, systems and workflows your ideas depend on, with the people who run them day to day.
You get · Readiness notes, with an owner for each gap
- 03
Ranked plan and review
We rank and cost the ideas, then write the roadmap and the guardrails. Your leads review the draft with us before you sign it off.
You get · A ranked, costed plan you have signed off
- 04Paid
Paid pilot
The engineers who wrote the plan build the highest-ranked idea in your systems, with a demo and a written update every week.
You get · A working pilot in your repo
- 05
Owner test and handover
Your team tests the pilot on real work. We hand over docs, runbooks and training, and write the board update.
You get · A board update backed by something that runs
Two places to stop: after the plan, or after the pilot. Everything written so far is yours, and you can take it to another team.
The shipped work behind our AI strategy consulting
HorizonLux has not published a standalone AI strategy engagement yet. The service grows out of the roadmap phase that opens every engagement we run, which ranks what to automate by hours saved and effort. Below are two systems we built and shipped, one free tool, and what each brings to your plan.
RepairCheck
AI damage inspection · GermanyVehicle damage inspection with AI damage analysis on OpenAI vision and a copilot that writes to the case record. Auto-estimates stop when an expert claims the case, and prompt settings and audit logs sit in the admin dashboard. For your plan: review points set by engineers who have built them into an AI product.
React Native · Expo · Next.js · Postgres · OpenAI · n8n
Read the RepairCheck case studyMasterPilot
Operations portalThe operations portal of a flight-training platform: webhook sync, a reporting engine, and weekly and per-flight email reports. 129 endpoints and 23 modules over 10 releases, live since March 2026. For your plan: effort scores that count the integration work, not only the AI part.
React · NestJS · Postgres · Prisma
Read the MasterPilot case studyAI cost diagnostic
Free toolEnter a workload and see an estimated monthly LLM bill, then which of five cost levers saves the most. We published the same method run on one of our own systems, line by line. For your plan: the running-cost row uses this method.
Free · no signup
Try the AI cost diagnostic
…how much ownership he takes: he thinks a feature through, raises edge cases before they turn into bugs, and delivers something that actually works… Exceptional clear and consistent communication.
Who should write your AI roadmap?
The usual options for a company without an in-house AI lead, side by side. Each one fits a different situation.
| What to weigh | HorizonLux | Large consultancy | Independent adviser | Your own team |
|---|---|---|---|---|
| Who writes the plan | Senior engineers who can then build the pilot | Strategy consultants; the build is often a separate team | One adviser, usually without a build team | Your leads, next to their day jobs |
| Where estimates come from | Your systems, checked hands-on, and systems we have shipped | Frameworks and benchmarks from many clients | The adviser's own past projects | Internal knowledge, with little outside comparison |
| Running cost per idea | Estimated for each ranked idea | Depends on the engagement | Depends on the adviser | Depends on who has run AI in production |
| What you have at the end | A written, ranked plan, plus a working pilot if you go ahead | Often a detailed report and presentation | A plan or workshop notes | An internal plan in your own words |
| After the plan | Optional paid pilot with the same engineers, or take the plan elsewhere | Usually a separate implementation project | You find a team to build it | Your team builds it, or you hire for it |
| Good fit when | You have 50 to 500 staff, no AI lead, and want one use case proven before a bigger spend | You need change across a large, multi-country organisation | You want a sounding board more than a build | You already have an AI lead with time to spare |
Based on how these options are usually offered, not on any named firm. Prices are not compared: ours is agreed on the free call, and we do not publish one.
Questions about AI strategy consulting
Straight answers on scope, price, time, ownership and what happens after the plan.
What should an AI strategy include?
An AI strategy should say where AI will and will not be used, in what order, with what data and controls, and how success is measured. A HorizonLux plan covers this in four written parts: readiness notes on your data and systems, a ranked shortlist with each idea's monthly running cost and the ideas to skip, an AI roadmap with owners and measures, and AI governance guardrails. The roadmap ends with the scope of one pilot, which you can build with us, with another team, or not at all.
How much does AI consulting cost?
HorizonLux does not publish a price for AI strategy consulting: it is agreed on the free 30-minute call, before any work starts. Scope sets the price: how many teams and ideas are in play, how many systems the readiness assessment covers, and how long the plan takes. The pilot is priced separately, before you commit to it. To see what the AI itself would cost to run each month, try the free HorizonLux AI cost diagnostic.
How long does AI strategy consulting take, and how does it start?
A HorizonLux plan for one team with a handful of ideas takes CONFIRM: typical plan length, for example two to four weeks. Several departments with dozens of ideas take longer. It starts with a free 30-minute call with Seif Sgayer, the founder, with no commitment, where scope, price and length are agreed. The pilot's length is written into its scope before it starts, and it brings a demo and a written update every week.
Is AI consulting worth it?
AI consulting is worth it when the plan changes what you build, and a poor buy when it ends in a deck nobody acts on. HorizonLux will say so if a plan is not worth paying for. If you already know your use case and your data is ready, skip the strategy and start with a build pilot. If you have many ideas, unclear data and a board asking for a plan, a ranking helps you avoid funding the wrong pilot.
Do we need clean data before we start?
No. HorizonLux checks your data as part of the work: the readiness assessment finds where each idea's data lives, who owns it and how usable it is. Some use cases, such as drafting replies from documents you already hold, can start with what exists. Others, such as demand forecasting, need the data fixed before a build. The plan says which is which, and what the fix would take.
Who owns the plan and the code?
You do. HorizonLux clients own every document, the code and the IP, and everything lives in the client's own repo. An NDA is available on request. If you stop after the plan, or want another team to build the next phase, you take the plan and the code with you. Nothing is held back to keep you tied to HorizonLux.
Which AI tools and models does HorizonLux recommend?
HorizonLux recommends the tools that fit the use case, your data rules and your budget. Its engineers build AI features on OpenAI and Claude models, and use Claude Code, Cursor, Codex, n8n and Supabase every day, so estimates come from hands-on use rather than vendor claims. The plan names the tools for each idea and explains the trade-off, so your own team can check the choice. Sometimes the answer is a plain rule with no model at all.
How does HorizonLux handle AI governance, privacy and the EU AI Act?
Governance is part of every HorizonLux plan, not an appendix. Each idea gets a risk score, and the guardrails set where a person must review AI output, which data each use case may touch, and a starter AI use policy. The NIST AI RMF and ISO/IEC 42001 are reference points only. An NDA is available on request. HorizonLux engineers are not lawyers: for EU AI Act duties they flag the questions, and your legal adviser decides.
Bring the board's AI question. Leave knowing where to start.
Thirty minutes with Seif Sgayer, our founder. Free, with no commitment.
Book a Free 30-Minute Call
Bring the board's question and the AI ideas on your list. Seif goes through them with you, points to the one or two that look ready for a pilot, names the likely blockers, and tells you plainly whether a full plan is worth paying for. If it is, you agree its scope, price and length on the call.
Don’t want a call? Email [email protected]
Book a free call- Your AI idea list, talked through with the founder
- A likely pilot, and what stands in its way
- Scope and price agreed before any work starts
We worked with Seif on the React Native app for our vehicle-appraisal platform. What stood out most was how much ownership he takes: he thinks a feature through, raises edge cases before they turn into bugs, and delivers something that actually works. And then there is one more thing why I wanted to work with Seif after our very first call: Exceptional clear and consistent communication. It is a pleasure to work with him and his team!
Michael EmaschowFounder, RepairCheck
Seif SgayerFounder ·View LinkedInFree AI Strategy Call
30 minutes · Google Meet · Free
No packages, no sales pitch. You leave with a clear first step.
- 30 min
- Google Meet
- Calendly
- No commitment
- The plan is yours to keep
- Built for teams at Tabeebi, RepairCheck and MasterPilot
HorizonLux is an independent company. OpenAI, Claude, Claude Code, Cursor, Codex, n8n and Supabase are trademarks of their owners. Naming them implies no partnership or endorsement.