You said it plainly on the call: there is no one who can compete. This plan is about the one place that is not yet true, which is capacity.

Every application costs a consultant's attention. That attention is your ceiling, and it is why your fees sit where they sit.

The last seven weeks were not about building a case tracker. They were about moving what lives in Riekus's head, and Claire's, and Brandon-Lee's, into something that runs on every file at once and does not get tired at four in the afternoon.

Two things happened in your own files while I was doing it. A renewal was refused while complying with the rule as the Department publishes it. And an appeal representation left your office citing the wrong section of the Act. Neither was carelessness. Both were caught by a person reading closely, which is exactly what does not scale.

What follows is what gets built, in what order, what has to be true for the dates to hold, and what is deliberately excluded.

A renewal in your book expired on 31 August. The pack was lodged on 1 July, sixty-one days ahead. The Department's published wording says no less than 60 days prior, making the last safe day 2 July. You were compliant with a day to spare.

It was refused under Regulation 9(8)(c). The Department had counted two calendar months back from expiry, which lands on 30 June. On that reading the pack was one day late.

Both facts were true at once. That is why the system never answers with a pass or a fail.

The gap runs from nought to two days depending on the months a case spans. It is computed per file, never assumed. Every renewal is scored against both readings, the earlier governs, and the handler sees both dates.

On top of that sits your own margin, because neither of you works to the legal minimum.

So the warning comes on your day, not the Department's last lawful one. A warning that turns red on the final legal day has arrived too late to act on. The statutory date is never overwritten. You and an auditor both have to see it.

An appeal representation in the submission pack opens on a visitor's visa renewal under section 11(1), and closes by asking the Department to issue a permit so the applicant can work under a General Work Visa in terms of section 19(2). Two different routes, one letter.

I raised it as a question, not a correction. Riekus answered the same week: the 19(2) is a mistake on the letter, yes. He judged the impact nil and the appeal is lodged. Nothing is owed on that file.

What matters is that this is the exact failure Rayne described in the first meeting, months before I found an instance of it:

It took someone reading sixty-eight pages to find it. From the build onward every generated letter has its citations checked against the route before release. A citation that does not belong stops the letter and quotes the sentence back.

I went through the platforms your competitors run and the firms you compete against. Section 11 has it. Every one can tell a client where a case is. Not one can say what goes wrong in the next six months, because none owns a model of South African immigration law. They own workflow software. You would own the rules.

This is the part that changes your pricing, and it is your idea rather than mine. When intake stops being limited by what one consultant can hold, the cost of serving an application falls. Fall far enough and you set a floor the firms in section 11 cannot follow, because they would have to rebuild to try. I have modelled it against your fee card and volumes and will walk you through it once you have read this.

Ten services, each doing one job, each able to be tested on its own. Described here by what they do rather than how they do it, for reasons section 12 sets out plainly.

Interfaces in and out. Sign-in runs against your corporate identity, Microsoft or Google, so nobody manages another password. Mail flows in and out threaded on the case. The document store sits behind one interface so it can be swapped without touching anything above it. An interface for corporate HR systems is designed in from the start and connected when a client asks for it.

Two decisions inside that shape are worth stating, because they are the ones a technical reviewer asks about first. The pages are built on the server and sent finished, with no separate application running in your browser, because the job is forms, tables and queues, and one codebase is easier to secure and to hand on. And documents are generated when somebody needs one, from the audited record, rather than stored: nothing is kept twice, any past version is reproduced from the history, and a submission record holds what actually went to the Department and what came back.

There is no Departmental interface, and any supplier who tells you otherwise is guessing. Nothing in this build assumes one. Qualification verification has no interface either, and each application sits behind the applicant's own mailbox, which is why that part is designed as a countdown you can see rather than an integration nobody can have.

One route at a time, each tested before the next begins, working back from full completion in December 2026. Each stage carries the test that decides it is finished. Those tests were written before the work started, which is the only way a sign-off is worth anything.

Behind all of it sits a workbook of 166 numbered requirements, each with a test written so it can be checked rather than argued about. Sixteen are built. Anything still marked unconfirmed waits for one of you. The workbook is yours to read at any point.

Two lists. The first is what I need from you. If one arrives late, the date moves. The second is what I have taken as true without you confirming it. If one is wrong, the work changes.

And what I have assumed. Say so now if any of these is wrong.

• Outbound visas are out of scope. It is effectively a second system and should be priced as one.
• Your consultants work on laptops, not phones, for case work. The client portal is built for phones; the staff console is not.
• Existing matters stay in your current process until you approve a move. Nothing migrates silently.
• The VAT split between Mobility and Solutions is carried through into every quotation the system produces. That both entities run under one agreement is settled, not assumed, and you settled it on 3 September.
• Your document templates remain yours. I build the machine that fills them, not a replacement house style.
• English only for this build.

You will read this and want things changed. That is the point of sending it. And once I am building you will think of something on a Tuesday that was not in the plan. Here is how that works.

Small things are free. Anything under half a day I absorb. Wording, a column in the wrong order, a field you want moved. No paperwork and no conversation about money.

Bigger things get one page. Anything above that gets a single page back from me within three working days: what it is, what it adds in days, what it moves in the schedule, and what it costs. Then you say yes, no, or later. Nothing is built until you have said yes in writing, and nothing is invoiced that you did not say yes to.

Everything agreed goes in a log that sits with the agreement, so in February neither of us is trying to remember what was decided in October.

No surprise invoice, and nothing built that you did not ask for.

Your clients are Kellogg, KLM, Air France, Huawei, Tata, TCS, Infosys and Mediclinic. Procurement at that level does not ask whether you have security. It sends a questionnaire and asks where the data sits, who holds the keys, and what happens when somebody uploads something they should not have. These are the answers.

Every control above is designed and specified. The document layer has been independently assessed and I hold the assessment. When your client's security team sends the questionnaire, the answers are already written.

And the question you are going to be asked this year: where does AI sit in this?

The platform's judgement is deterministic and stays that way; that is what makes it defensible in front of the Department and your clients' auditors. On top of it sits an assistance layer that removes typing and catches mismatches: it reads a passport and asks the person to confirm, it audits a pack against the checklist and the route, it names any known refusal ground that applies, and later it will answer status questions from the record and draft the motivation paragraph a consultant then edits and signs. It never writes a date, a score or a section of the Act, every suggestion is confirmed by a person, every call is logged and metered, and nothing about your applicants leaves South Africa.

None of that layer is in this price, and none of it is switched on for you until it has been shown to work on a set of your own cases where the right answer is already known. You are hearing it from me rather than from a competitor's brochure because you are going to be asked, and the honest answer is more useful to you than an impressive one.

Also deliberately out of scope: anything that automates the submission itself; any model writing letter content or touching a score or a date; the qualification portal, because there is no interface to it and each application sits behind the applicant's own mailbox; any public-holiday calendar I have not been given a source for; outbound visas; and the radar in section 04, which is phase two.

The firms that surface against corporate work-visa searches in South Africa are IBN Immigration Solutions, Xpatweb, IMCOSA, Black Pen Immigration, Work Permit South Africa and Hassle Free Immigration. Black Pen holds a boutique corporate immigration award. Xpatweb markets a success rate. IBN markets its awards. They compete on reputation and on relationships, which is the ground you compete on and the ground you already say nobody can take from you.

I also went through the case-management and mobility platforms firms at this level run on, here and internationally. What matters is not which badge sits on the software. It is what the category can and cannot do.

Everything I found is status software. It records where a case has got to and tells a client the same. I went looking for the two things this build leads with. A scoring hold on a mandatory criterion, and a deadline engine that understands the date-of-entry trap. Neither appears in any product marketing I could find.

That is the gap. Reputation gets you considered. Capacity, and the price floor that comes out of capacity, stops the conversation happening at all.

And it compounds where it matters most, which is your clients. A mobility manager at a multinational does not change supplier over a nicer portal. They change when an assignee is refused, when a day budget runs out unnoticed, when a renewal is lodged inside a window nobody flagged. Every one of those is a thing this system sees before it happens. The firms in this section will still be answering where a case is. You will be telling a client which six of their thirty secondments will be refused as packaged, and what to change. That is not a better service. It is a different conversation, and it is very hard to leave.

Your numbers, not mine. Two of them decide whether this is worth doing.

Your fee card gives a blended professional fee of about R7,350 an application. That weights core work visas at R10,500, the 11(1) family at R3,000, the 11(6) and 11(2) tier at around R4,000, and permanent residence at R15,000. Against the volume you gave me, that is the table below.

At forty a month, a fifty percent gain is R1,764,000 more in professional fees a year. On a thirty percent margin that is R529,200 of profit you are not making today.

That gain is not people working faster. It is the ceiling moving. Today the number of applications you can carry is set by how many one consultant can hold in their head. That is why an influx hurts, and why a holiday hurts. Move the deadlines, the scoring and the document checks out of memory and the same team carries more without anyone working later.

What it costs to run, and who carries it.

Named now rather than later.

One thing to say now rather than when it becomes relevant. Those figures are the floor for one environment, and they are a floor rather than a usage charge: the smallest database and the smallest service still cost that at zero volume. So a second environment, an airline running its own instance for example, carries its own floor from the day it exists, whether or not anybody has used it yet. I would rather you heard that now than at the point somebody asks for one.

No licence fee. No charge per seat. The infrastructure sits in your own cloud account in the South African region. You see the bill, and it goes with you.

Two costs sit with you rather than me. A penetration test before go-live, which procurement at Kellogg or KLM will ask to see. And your attorney's data-protection position, named in section 10. Neither is marked up by me.

My fee and the per-application arrangement are not in this document. They belong in the commercial agreement, which follows your sign-off on the scope in section 06. Agree what gets built first. The money is a shorter conversation once that is settled.

Everything above is one build for one firm. It was not designed as one.

There is a deliberate seam in it. The rules sit apart from the machinery that runs them. That is why the build refuses to guess when a calendar is missing, and why a rule change is a versioned edit rather than a rewrite. It is also why this can go further than South Africa. The machinery does not know which country it is in. Only the rules do.

I have routes of my own that this serves. The United Kingdom, where I work already and where corporate immigration has the same shape and the same refusal-on-a-technicality problem. Latin America, where I have an opening I have been holding rather than spending, because until now I had nothing worth walking in with. Those are mine to develop, and where the platform earns from them I would rather that ran through an arrangement with you than around you.

And you are not one company. Nash Group Mobility sits inside a group, and the parts of this that are not immigration-specific are not specific to visas either. The custody of sensitive documents. The deadline discipline. The forecasting. The client portal. Where any of that is useful to another business under your name, the work is already paid for.

And when you say you would give it to your customers, that is two different things with two very different price tags, so it is worth separating them now. The first is already in this build: an airline's mobility team signs in to your platform as a corporate client and sees only their own people. That is the client portal in Stage 3: no new environment, no new brand, nothing extra to buy. The second is an airline running its own instance, with their branding and their domain and possibly their own consultants. That is a separate project with its own environment and its own fee, and it is priced when you want it. The architecture is built so the second one stays cheap: one codebase for everyone, with what differs per client held in a configuration file rather than in code.

None of that is in this price and none of it is in this scope. It is here because a build plan that ends at handover is a transaction, and I am not proposing a transaction. The first thing we ship together is a system that stops your firm losing visas on a two-day gap. What it becomes after that is a conversation I would like us to be having in a year.