Custom AI solutionsFor 30-400-person SMEs

One bottleneck.
One solution.
Live in 10 days.

Don't know where to start with AI? We pick the bottleneck with you, build the solution, and ship it into your stack in 10 working days.

Live in 10 days. Don't pay until it works.

Not a refund. Not a credit. If it's not in your stack saving hours by day 10, you owe us nothing.

No-cost diagnosis. 30 minutes. Yours to keep.

Works with the tools you already useYour data stays in your stackNo better time to start
Pick the one eating your week

Don't see yours? Tell us in the scan. 80% of what we ship is custom.

Lives in the tools you already use

QuickBooksXeroZohoSAP Business OneMicrosoft 365Google WorkspaceSlackWhatsApp BusinessOr your custom internal tools
Why most AI projects fail in SMEs

Three reasons it didn't work last time.

You've been pitched by 22-year-olds with Make.com automations. We've watched it go nowhere too.

01

Generic AI on a specific business

ChatGPT wrappers and "AI strategy decks" don't know your suppliers, your VAT rules, or how Hassan actually files invoices on Tuesdays.

02

6-month roadmaps for 2-week problems

You don't need a transformation. You need one agent shipping work by next month.

03

Pilots that never go live

The demo looked great. It still doesn't touch a real invoice. We don't do pilots, we ship into your stack.

How we deliver

Scan. Build. Run. Ten working days.

No retainer fishing. No "AI strategy" deck. We pick one bottleneck, ship one solution, and prove it before you pay.

1

Scan. 3 days

A 30-min call, then we shadow the work for a few hours. You get a written diagnosis: where the hours leak, whether AI is actually the right fix, and what the solution would do.

2

Build. 7 days

We build the solution against your real data, on your stack. The best generative model per use case for the reasoning, orchestrated through your existing tools.

3

Run. Day 10 onward

Live in your stack. We monitor, fix, and tune for 30 days. After that, you own it, or we keep running it for a flat monthly.

Honest note on who builds it

Depending on the use case, the build is handled either by us or by a small implementation partner we bring in. We run the scan, the spec, the QA, and the handover. If you want your own team to learn and start improving these systems over time, we can support that too.

Live agent · this is real

Watch one of the solutions work.

Two patterns we ship most often. Click through to see what each one looks like once it's running on real data.

Live · processing supplier invoices

The Invoice Agent

Live

Watches the AP inbox + WhatsApp, extracts line items from PDFs, images, and scans, matches POs, posts to your accounting system. Flags anything weird for a human to review.

94%
Auto-posted
2.1s
Per invoice
5
Formats handled
INV-AM-4821Al Manzil · 2 min ago
PostedPO-2812 ✓
IMG_0429.jpgWhatsApp Hassan · 4 min ago
ExtractedAwaiting GL
Scan_03142.pdfEmail F&B Supplies · 7 min ago
PostedPO-2814 ✓
ar-فاتورة.pdfEmail Arabic · 11 min ago
PostedPO-2811 ✓
IMG_0431.jpgWhatsApp Reem · 14 min ago
FlaggedNo PO match
What the first build looks like

We're new. So we're showing you the target, not a fake case study.

These are the outcomes we're committing to with the first three builds. Pulled from agent builds we've scoped, costed, and benchmarked against shipped agents in the same stack.

We'll publish the real numbers, with named clients, the moment they're live.
Hospitality · target
Group F&B operator
Bottleneck: Supplier invoice processing
Before
16 hrs/week
across 2 AP staff
Target
90 min/week
one reviewer · 94% auto
Services · target
Mid-size services firm
Bottleneck: Internal SOP & policy questions
Before
8 hrs/day
across the team, asking "the one person who knows"
Target
Under 1 hr/day
Slack assistant resolves 91% with citations
30-400 staffArabic + EnglishWhatsApp-firstLegacy ERP

Built for 30-400-person SMEs.Multi-supplier, multi-channel, multi-language. If your operation lives in WhatsApp, Excel, or a 2014 ERP, we know the terrain. Sectors we've productised are listed below; the rest get a custom build.

Verticals

When a pattern repeats, we productise it.

We build custom solutions for SMEs. When the same problem shows up across a sector, we turn it into a vertical SaaS tool. Real estate is the first.

F&B, distribution, and logistics next. Tell us your sector.

Cost of waiting

What's the bottleneck costing you each month?

Move the sliders to your real numbers. We're not choosing them for you.

Hours per week per person on this task20 hrs

For mid-size hospitality AP teams, this is usually 15–25.

People doing it2

Don't forget the manager who reviews and chases.

Loaded hourly cost (€)35/hr

Salary + overhead + the cost of mistakes.

What this bottleneck costs you
€72,800 / year

€6,062 / month · 40 hours/week of operator time

An agent at this volume typically claws back 70-80% of the cost, plus the things harder to put a number on: more consistency, fewer errors, and time freed up to solve deeper problems.

Weekly
€1,400
Monthly
€6,062
Yearly
€72,800
Get my Bottleneck Scan
Skip the form

Pick a slot. We'll see you on the call.

30-minute scan. Your real workflow, our real questions, written diagnosis 2 days later.

Pricing

€5,000 to build it. Monthly to run it.

Each solution is built custom around your workflow. €5,000 one-off for the build, then a monthly subscription per solution to run, host, and tune. Tell us your niche.

Custom solution

One bottleneck. One solution. Live in 10 days.

Built around your real data and your stack. Yours to use, ours to maintain.

Bottleneck Scan
Half-day shadow + written diagnosis. Yours to keep.
Day 1-3
Solution build
Custom-built into your stack, on your real data.
Day 4-10
Live + 30-day tuning
We monitor, fix, and tune. You don't lift a finger.
Day 10-40
Custom build
5,000

One-off build. Then from €400 per month per solution to run, host, and tune.

Cancel anytime. Already-productised verticals (like real estate) skip the build fee and run on per-number pricing.

Tell us your niche
If it doesn't go live in 10 working days, you don't pay.

Not a refund. Not a credit. You simply don't pay. We eat the build cost.

About Bottleneck AI

Built by people who use AI agents in their own work, not just talk about them.

Three of us: one AI agent engineer, one industrial engineer specialised in AI, and one business operations lead. Technical depth on one side, commercial and operations on the other, shipping production solutions inside SMEs.

We believe AI will be central to operational automation over the next few years, but only if you actually integrate it. Pasting prompts into ChatGPT isn't enough. The real value shows up when an agent becomes part of how the work actually gets done: reading the inbox, posting to your accounting system, answering staff in Slack, sourcing prospects from your own database.

The pattern that kept winning across 200+ use cases we evaluated was small, sharp, and boring: one bottleneck, one solution, into the existing stack. Multi-agent orchestration, platform plays, "AI-first transformation" decks were theatre.

Want to know who you'd be working with? We walk you through the team and, where relevant, the delivery partner on the scan call. If your team wants to own more of the build over time, we can support that as well.
Real questions, straight answers

Six things buyers actually ask.

What does the €5,000 actually cover?
The complete build of one custom solution, deployed into your stack and live on your data. That includes the half-day Scan, the spec, all the engineering and integration, and 30 days of tuning after go-live. Then a monthly subscription per solution to run, host, and tune (from €400/mo depending on scope).
What if the solution doesn't go live in 10 days?
You don't pay. We eat the build cost. The €5,000 is conditional on shipping. If we miss, that's our problem, not yours.
Where does our data live?
Inside your existing tools. The solution reads from and writes to the systems you already pay for: your accounting, your CRM, your email, your WhatsApp Business inbox. We don't centralise data into a Bottleneck AI database. For the AI itself, we pick the best generative model per use case (open source, Claude, or OpenAI) and check the model's compliance posture against the data it'll touch before we ship.
Our use case isn't in your demo list. Can you still help?
Probably yes. The two demos are the patterns we see most often, but ~80% of what we ship is custom. Tell us your niche on the call. We'll tell you in 30 minutes whether it's a fit, before either of us commits.
Last call

Stop trying to "do AI." Tell us the bottleneck. Leave the rest to us.

30 minutes. Written diagnosis. No deck. If we're not the right team, we'll say so on the call.

Custom solutions for SMEs. €5,000 + monthly. Real estate vertical from €500/number/month.