AI solutions · The flagship
One operations bottleneck.One production AI tool.Ten business days.
Fixed price. Fixed date. The code lives in your GitHub from day one and is yours outright on final payment.
A demo every week · The clock pauses only when we wait on you
Week 1
Week 2
01The bottleneck
Your team is the integration layer.
In most forwarding, brokerage and 3PL offices, skilled people spend the day moving data between documents, inboxes and the TMS. That is the work we take off their desks, one workflow at a time.
01
Documents typed by hand
Today
Bills of lading, commercial invoices, packing lists and customs declarations arrive as PDFs, scans and phone photos, in Arabic and English. Someone reads each one and keys it into the TMS, field by field, then someone else checks it.
The fix
An intake agent reads each document in any language, extracts the fields you actually use, cross-checks them against each other and posts them to your system. Anything uncertain goes to a review queue with the source highlighted. A person approves; nothing is posted blind.
02
Quotes that wait for one person
Today
A rate request comes in. Someone opens three spreadsheets and a carrier portal, digs through last month’s email, and builds the reply by hand. The customer asked two other forwarders at the same time.
The fix
A quote agent reads the request, pulls lanes and rates from your sheets or system, and drafts the reply in the customer’s language. Your team checks the numbers and presses send. The margin logic stays yours.
03
Inboxes and WhatsApp threads
Today
Shipment updates, document requests and customer questions land across shared inboxes and a dozen WhatsApp groups. Things slip because nobody owns the thread, and the context lives in someone’s head.
The fix
A triage agent classifies every message, links it to the shipment, drafts the routine replies and routes the rest to the right person with the history attached. Replies and review screens work in your customers’ languages. Every action is logged.
02Builds
The Builds menu.Fixed price, fixed clock.
Each Build is one production AI workflow shipped into your real systems, quoted as one fixed number before it starts. Pick from the menu or bring your own bottleneck; a Blueprint tells you which one pays back first.
Document IntakeLogistics
Document Intake
Bills of lading, invoices, packing lists and customs documents, in any language, Arabic scans and phone photos included, into your TMS, ERP or sheets, with a human review queue.
Any language
- Clock
- 10 business days
StandardLogistics
Inbox Triage
Email and WhatsApp (the official Business Platform) classified, linked to the shipment and routed, with routine replies drafted.
Any language
- Clock
- 10 business days
StandardLogistics
Quote Drafting
After Inbox Triage: rate lookup, drafted replies in your customer’s language, an approval step and a quote log.
Any language
- Clock
- 10 business days
StandardLogistics
Ops Copilot
Answers over your SOPs, documents and shipment data, with permissions and a citation for every answer.
- Clock
- 10 business days
StandardWordPress
WordPress AI feature
An AI capability inside your plugin, theme or WooCommerce store, shipped the way your users install things.
- Clock
- 10 business days
StandardSaaS
AI integration into your SaaS
One agent, retrieval pipeline or AI feature taken all the way to production in your product.
- Clock
- 10 business days
- Tell us what you need
Not on the menu?
If it is one workflow and it fits the clock, it is a Build. Tell us the bottleneck and we will tell you, plainly, whether it fits.
03Every Build includes
Governed production, as standard.
A demo is the easy part. This is what makes a tool safe to leave running on a Monday morning, and it comes with every Build.
Authentication and permissions
Your users and roles. The tool only sees what the person using it is allowed to see.
Audit log
Every input, output and human approval recorded and searchable.
Monitoring and cost dashboard
Errors, latency and model spend per workflow, visible from the first demo.
Evaluation set
Real examples with known answers. Every change is scored against it before it ships.
Automated tests and CI
The pipeline runs on every commit, in your repository.
Handover docs and walkthrough
Written runbooks and a recorded walkthrough for your team.
30 days of fixes
If it breaks after demo day, we fix it.
Your repository, day one
Created in your GitHub organisation before the first commit. Yours outright on final payment.
04Working together
Blueprint, Build, Run.
Three ways to work together on AI. Each one starts with a written fixed quote; no hourly billing and no surprise invoices.
Blueprint
01/03One department or workflow, mapped and proven before you commit to a Build.
- Clock
- 5 business days
- Two or three recorded discovery calls
- A process map of the workflow as it runs today
- A scored roadmap: impact against effort, hours and documents saved
- A production architecture sketch: permissions, audit log, data residency
- A working demo of the top fix, on your own sample documents or emails
The full Blueprint fee is credited to a Build signed within 30 days.
Start with a BlueprintTen-Day Build
02/03One production AI workflow shipped into your real systems, on a fixed date.
- Clock
- 10 business days
- A numbered scope and a demo date agreed at signature
- The repository in your GitHub organisation from day one
- A working demo every week, on your data
- Everything in the governed-production checklist
- 30 days of fixes after demo day
Anything not in the numbered scope becomes the next Build, never a surprise invoice.
Tell us what you needRun
03/03We keep it working, keep it affordable, and keep making it better.
- Clock
- Ongoing
- Monitoring, model and prompt upgrades
- Evaluation re-runs on every change
- Cost control
- 2 shipped improvements a month
- Next-business-day response, 2-business-day fix target
- A monthly usage and ROI report
Model usage runs on your own API keys, billed to you directly by the provider.
Ask about Run
05Free, no call required
Show us your documents.
Send 10 redacted shipping documents. Within 48 hours you get a working extractor running on your own paperwork.
Documents in any language work, Arabic scans and phone photos included.
Email your documentsSubject line prefilled: “Show us your documents”
- Documents
- 10
- Turnaround
- 48h
- Slots
- 5/mo
01You send
You send. 10 redacted bills of lading, commercial invoices or packing lists. Mixed layouts, bad scans and phone photos are welcome; that is the point.
02We build
We build. An extractor for the fields your team keys in today, run against every document you sent.
03You get
You get. The working extractor, a field-by-field accuracy score, and a short video walkthrough of the results.
Free. No call required. Limited to 5 companies a month. Send only documents you have the right to share. The sample documents on this site are synthetic; yours are deleted after 30 days.
06AI questions
What operations leaders ask us.
Is ten business days realistic?
For one workflow with a numbered scope, yes. That is why a Build is one workflow and not a platform. If a request does not fit the clock, we say so on the first call and suggest a Blueprint, or two Builds in a row.
Where is my data processed?
The tool and its data run in your cloud or region of choice. The model provider and processing region are named in the data processing agreement. In-region model processing for the Gulf and the EU is confirmed per project before we promise it.
Which models do you use?
Models from Anthropic (Claude) or OpenAI, chosen per task and measured against your evaluation set. Usage runs on your own API keys, billed to you directly by the provider.
How do you know it works?
Every Build has an evaluation set: real examples with known answers, taken from your own documents or messages. Every change is scored against it before it ships, and you see the scores.
What happens when the model changes?
We re-run the evaluation set on the new model before switching. Under Run, upgrades and re-runs are included. Without Run, you have the evaluation set and the docs to do it yourself.
Do you automate WhatsApp?
Only through the official WhatsApp Business Platform, on your own verified business account. Business verification must finish before the clock starts. We never automate WhatsApp Web or personal numbers.
Something else? hello@granitebolt.com
Bring one bottleneck.
30 minutes. We tell you whether it fits in 10 business days and what it costs. If it does not fit, we say so.
Or write to hello@granitebolt.com
Your time and ours
Calls seven days a week
Beirut
Dubai
Riyadh
Doha
Kuwait City
Manama
Muscat
Cairo
Berlin
London
New York
Beirut shares the working day with Dubai, Riyadh, Cairo, Berlin and London, and the afternoon overlaps the New York morning.
One working day, several time zones.