You already know the one thing AI should fix in your business. We start there.
Veld audits, builds, and runs the data systems AI needs to work. Most AI stalls because the data underneath is trapped in separate tools. We fix that — then keep it running.
Tell us the problem eating your week — the spreadsheet, the re-typing, the 9 a.m. scramble. We’ll tell you, in hard numbers, whether AI can fix it — and what it won’t.
A sample of what we find, and what we build.
Drawn from real engagements. Anonymised at the client’s request; the words and the numbers are theirs.
lead times
every day
nightly
readiness
One case study is named and detailed below. The rest stay anonymous — the pattern is the point.
Where does your week actually go?
Pick the one that stings most. We’ll show you what AI does well there, what it won’t touch — and the catch nobody mentions.
Not on the list? It almost certainly rhymes with one of these. Tell us what’s breaking and we’ll say, plainly, whether AI is the answer.
The catch is the same one. Every time.
You came in with one problem. The reason it persists — and the reason a bolt-on AI tool breaks within weeks — is the plumbing underneath: your data sits trapped in separate tools, never cleaned, never connected.
Point AI at a clean, connected data layer and the problem you named goes away. So do three you hadn’t named yet. That is the whole difference between an AI demo that wows the room and AI that quietly earns its keep, every month, for years. It is also why we never start by selling you a tool.
First we make sure we’re solving the right problem.
Before we build anything, a short, fixed-price audit points AI at the thing that’s actually costing you, puts a number on the fix, and tells you honestly when the answer is “not yet”. Ten days. Five phases. About five hours of your team’s time — the rest is ours.
Your AI readiness score
Six dimensions, one honest number, on a one-page executive summary.
Three high-ROI use cases
Ranked by what they are worth in your business. In hard numbers, not abstractions.
ROI projections on your numbers
Admin time recovered, cycles shortened, capacity gained. Auditable, lender-ready.
Process & data audit
Every system you run, with an honest picture of what is ready to automate and what needs cleanup first.
An interactive demo build
A working, clickable version of your future system — generated from the audit, hosted on your branded domain.
A 90-minute live walkthrough
A working session, built for questions. The report is yours either way.
▌ Process structure
▌ Technology stack
▌ Team capability
▌ Commercial opportunity
▌ Risk & compliance
Every report we deliver says “no” somewhere.
The most common way AI spend goes to waste is automating the wrong thing. So every audit draws the line in writing — a real example, straight from a delivered report.
- Extracted structured data to pre-fill application forms and invoices
- Cross-referenced documents and flagged mismatched fields for review
- Flagged illegible inputs at intake, before they reached production
- Edge-case classifications that carry regulatory consequences
- The human sign-off layer — identity and serial verification stayed human
- The founder’s relationships and judgement calls
We’re not here to make anyone redundant. The pattern across every engagement: capacity per person goes up, judgement stays human, and headcount grows because the business grew — not because the software failed.
We don’t hand you a fix and leave. We run it.
The reason most AI wins quietly stop working: APIs change, data sources drift, the business grows — and a one-off build has nobody minding it. So once we’ve solved your problem, we stay. We stand up the data layer, then maintain and grow it as a living asset, month after month. One price, the full stack included, one relationship — which is exactly why the problem you came in with doesn’t creep back.
The data layer, not the dashboard
Front-ends can be built in hours and have no moat — so they’re project-priced or yours to own. The value is below: extracting from Shopify, Meta and Google Ads, loading it, modelling it, and maintaining the semantic layer that makes your data usable by an LLM.
Skin in the game
A subscription rewards a clean, healthy, low-touch system — not manufactured hours. We want the asset compounding every month, because that is what makes the relationship endure. We are a partner, not a vendor.
Below the cost of one junior
Senior data engineering and the full software stack, for less than a single junior data engineer in your market would cost. No separate invoices for tools or infrastructure. No hidden costs.
The front-end
Reports, dashboards, portals. Impressive, fast to build, and easy to replace. You pay for it once, own it, or upskill someone to run it. It is not what the subscription is for.
The managed data layer
Extract, load, transform (ELT) on a standard stack — Airflow, Postgres/Snowflake, DBT — plus the semantic layer underneath. This is the work that can’t be casually replaced. This is the living asset.
Bagel Media — Daily Bagel.
The founder spent five hours a day on content production. Cross-platform performance lived in four dashboards that didn’t agree. The team used AI daily — individually, with nothing shared and nothing compounding. Every session started from scratch.
Ready to build. The half-points lost were the finding: capable team, individual habits, no organisational infrastructure. A 22-page report, built to circulate.
Two tracks, sequenced and costed monthly. Week one: shared organisational AI practice on the tools already paid for. In parallel: a unified data foundation, a single live view of the business, and content systems that compound. Budgeted to the cent against a stated set of assumptions — auditable, not hand-waved.
“Veld has been the reason we’re able to perform like a 20-person media business with a headcount of only 5.”
Three ways in. All of them priced, in writing.
No discovery-call reveal. No “it depends.” Most clients start with the audit — it’s the clearest picture of what’s worth building, and what isn’t.
Fixed fee · 10 business days · the full six deliverables, including the interactive demo build · 50% deposit, 50% on delivery · 7-day terms · card or bank transfer. The fastest way to know — in hard numbers — what AI is worth in your business, and what it isn’t.
Already know the one thing that’s costing you? We scope that single use case on the discovery call and come back with a fixed quote — fixed price, fixed outcome, no full audit required. The lightest way in when the problem is clear and contained.
Once you’re ready to build, we stand up and manage the data layer — stack included. A one-time build/onboarding fee covers the setup; the monthly is pure ongoing value. Six-month minimum.
Owner-led businesses, 5–100 staff, that run on data — e-commerce, logistics, media, professional services — and already have one problem they believe AI can fix. If that’s you, the call will be worth your 30 minutes. If it isn’t, we’ll say so.
per month
Every engagement begins with a 30-minute discovery call. No charge, no pitch — we decide together whether to go ahead.
The questions everyone asks. Answered in writing.
“Can you just fix one specific thing for us?”
Yes — that’s often the fastest way to work with us. If the problem is clear and contained, we scope that single use case and quote it at a fixed price, no full audit required. Often, fixing it well surfaces the deeper opportunity — and if it does, we’ll tell you plainly and you decide whether to go further. No obligation to.
“We’ve been sold ‘automation’ before. It didn’t work.”
So were most of our clients — half-working integrations and vendor promises are usually why we get called. The audit is a fixed-price diagnostic that stands on its own; the report is yours either way. If we go further, it’s as a partner managing a living asset, with our incentives tied to keeping your system clean — not to selling you more.
“Why is a subscription better than just paying for Claude or ChatGPT?”
Because those tools run on your data plumbing, and that’s what we build and maintain. A chatbot can’t extract from Shopify, load and model your data, or keep a semantic layer healthy as your APIs change. We’ve told clients to keep tools they already pay for and consolidate rather than add — the value is the layer underneath.
“Why a subscription and not a one-off project?”
Software is never done. APIs change, data sources evolve, the business grows. A project promises a finish line that these systems don’t have — they take about three months just to stabilise. A subscription reflects the reality: a living asset that needs care, with a price that sits below the cost of one junior engineer.
“Will AI replace my people?”
We’re not here to make anyone’s role redundant. The pattern in our work: each person handles more, judgement stays human, and the business stops needing to hire linearly just to keep up with admin.
“Is our data safe with you?”
NDA before any system access. Read-only access or screen-share only. And your data-protection position is assessed inside the report itself — it’s one of the six dimensions, not an upsell.
“How much of my time does the audit actually take?”
About five hours, total, across the engagement: one 60-minute discovery session, a few 20–30 minute calls with your operational leads, and a 90-minute walkthrough. Everything else is handled by Veld.
The people who run your audit are the people who build it.
No account managers, no juniors, no handoffs. A small senior team takes you from audit to running system and stays on it — which is how the incentives stay honest and the knowledge never leaks.
Veld — Afrikaans for the open field. Full of structure, for those who know what to look for.
Tell us what’s breaking.
Describe the problem in a line or two. We’ll tell you honestly whether AI can solve it — and whether we’re the right people to do it. 30 minutes, no charge, no pitch. We reply within two business days.
- A 30-minute call — no charge, no pitch
- We reply within two business days
- NDA before any system access
- The audit report is yours to keep, either way
Every engagement begins with the problem, not a pitch. We decide together whether to go ahead.
What we collect, and where it goes.
This is a plain-English, privacy-aware notice. Veld runs two forms on this site: a discovery-call form and a newsletter sign-up. Each collects only what you type into it.
Form submissions are sent by email to philip@vldt.co.za (and, where configured, to a connected sheet) so we can reply to you. We do not sell your information or share it with third parties beyond the tools needed to deliver our reply.
We do not set advertising cookies. Any site analytics, if present, are anonymous and aggregated.
To request access to, correction of, or deletion of your information, email philip@vldt.co.za and we’ll action it.