AutoHive Business Consulting organises its work using Porter's value chain (Porter, 1985), which divides a firm into primary activities that create value directly and support activities that make the primary ones possible. The model has five numbered layers. Governance is firm infrastructure, the ground everything stands on, covering structures, trusts, constitutions, succession, the board and financial oversight. Marketing and sales is a primary activity covering demand, qualification and conversion. Operations and data is a primary activity covering delivery, evidence and the audit trail. The people lifecycle is human resource management, a support activity covering hiring, onboarding, training and representation. AI infrastructure is the fifth layer and the rails the rest runs on: the website, the CRM and the dashboards, which every other bucket reports into, so that the business holds one version of the truth rather than four. Above all of it sits the digital home. Your website is the home you own and hold the deed to, while social channels, marketplaces and ad platforms are digital Airbnbs you only rent, so every rented channel must push people and their data back into the home and the CRM. The sequence is governance first, then operations, then demand, with people running through all of it, because automation multiplies whatever structure it is built on.
The value chain model
Governance is not one service among several. It is the ground the others stand on.
A business that automates on top of a broken structure automates the breakage. If nobody has agreed who owns a decision, an automated workflow will not settle the question, it will simply produce the wrong answer faster and in greater volume, with an audit trail proving you did it on purpose. This page sets out the model that organises the whole division, and shows where every piece of work sits inside it.
The academic spine
Porter split a firm in two. We use the same split.
Michael Porter (1985) described a business as a chain of activities rather than a single lump. He divided them into primary activities, which create value the customer can see and pay for, and support activities, which make the primary ones possible without appearing on the invoice. Nobody buys your filing system. Everything you sell depends on it.
The four buckets in this division map onto that split without any stretching. Governance is firm infrastructure, the support activity that every other activity leans on: legal form, ownership, decision rights, oversight and the financial picture. Marketing and sales is a primary activity, the part of the chain that creates and converts demand. Operations is a primary activity, the part that actually delivers the thing and produces the record proving it was delivered. People, or human resource management in Porter's own wording, is a support activity, and like all support activities it touches every primary one.
That mapping is what makes these four a model rather than a menu. A menu lets you order the item that appeals. A model tells you what an item depends on, and what breaks if you take it out of order. When someone asks whether they can skip the governance work and go straight to automating quotes, the honest answer is not a sales position, it is a structural one: quotes are a primary activity and they run on firm infrastructure, so the infrastructure gets built either deliberately or by accident.
The model, drawn
One ground, one set of rails, three activities running on them.
Text equivalent of the diagram. The model is a stack of five numbered layers plus a band at the top showing where the customer arrives. Read it from the bottom up, because each layer rests on the one beneath it.
At the very bottom, running the full width, is layer one, Governance, the ground everything stands on. It carries the structures, trusts, constitutions, succession arrangements, the board, and the financial agents that watch the numbers. In Porter's terms it is firm infrastructure, a support activity.
Directly above governance is layer five, AI infrastructure, the rails, also running the full width. It is the website, the CRM and the dashboards that every bucket reports into. It is numbered five because it is built last, and drawn low because everything above it runs on it.
Standing on those rails are three hexagonal buckets, read left to right. Layer two, Marketing and Sales, covers demand, qualification and conversion, and is a primary activity. Layer three, Operations and Data, covers delivery, evidence and the audit trail, and is also a primary activity. Layer four, People Lifecycle, covers hiring, onboarding, training and representation, and is a support activity, human resource management in Porter's wording. A short arrow runs down from each bucket into the AI infrastructure layer.
Across the top, under the label where the customer arrives, sits the digital home band. It is not numbered, because it is the place people turn up rather than a bucket of work. On the left is a house shape labelled your digital home, the website, marked owned. To its right are two dashed boxes marked rented: social channels, described as digital Airbnbs, and marketplaces and ad platforms. A note beside them reads that rented ground can change the rules, or evict you. An arrow runs down from this band into the buckets below.
- One. Governance. Firm infrastructure, the ground everything stands on.
- Two. Marketing and sales. A primary activity, and the one that routes into the marketing division.
- Three. Operations and data. A primary activity, built and run by this division.
- Four. People lifecycle. Human resource management, a support activity running through all the others.
- Five. AI infrastructure. The rails: the website, the CRM and the dashboards.
- The digital home band, unnumbered. Where the customer arrives, not a bucket of work.
The picture language, in plain words
The diagram leans on a few figures of speech. If any of them did not land, here is what each one is actually saying. Nothing is lost by reading the literal version.
- Your digital home means your website
- A house is drawn because a website behaves like property you own. You hold the deed, you decide what goes in it, nobody can change the locks, and whatever value you build into it stays yours. It is the only place in your online presence where that is true.
- Digital Airbnbs means social channels and marketplaces
- You furnish the room and you bring the guests, but you do not own the building. The platform sets the house rules, changes them without asking, decides who sees your posts, and can close your account tomorrow with no notice and no appeal. Renting is not a mistake. Renting while owning nothing is.
- The rails means your website, CRM and dashboards
- Track has to be laid before a train can run on it, and everything afterwards is limited by where the track goes. Your systems work the same way. Every part of the business runs on them, so their reach sets the limit of what the business can do.
- The ground everything stands on means governance
- Foundations are invisible once a building is up, and nobody admires them. They also decide how many floors you can safely add. Governance is the same: unglamorous, largely unseen, and the thing that determines how much you can build before something gives way.
- Buckets means groups of related work
- Simply a way of sorting the work into four sets so it can be discussed one set at a time. No deeper meaning is intended.
The two layers people skip
The home you own, and the rails that carry everything.
Your website is the digital home you own. Everything else is an Airbnb.
You hold the deed to your website. Nobody else can change its rules, throttle its reach, decide who gets to see it, or close it while you are asleep. It is the one place on the internet where the terms are yours, and it is the only channel where an improvement you make this year is still yours in five years' time.
Social channels, marketplaces and ad platforms are digital Airbnbs. You are renting. The listing rules can change overnight, the algorithm can quiet you without telling you why, and the account can be suspended by someone you cannot phone. The room is comfortable and well located, but you are a guest in it, and guests get asked to leave.
Rent them deliberately, because that is where people already are. A business that refuses to show up on the platforms its market lives on is not being principled, it is being invisible. The rule is narrower than that: never let the rented rooms hold the only copy of your customer relationships. Every rented channel should push people, and the data that comes with them, back to the home you own and into the CRM, so that the relationship survives the platform.
Say plainly what this means in practice: an audience that lives only on a social platform is not an asset you own, it is an asset you are borrowing. It cannot be valued in a sale, it cannot be relied on in a downturn, and it can be withdrawn by a party who owes you nothing. A list of contacts sitting in your own CRM, gathered through your own site, is an asset. Forty thousand followers on a platform you do not control is an audience, which is a different and far more fragile thing.
Building the home is what the website development division does. Filling the rented rooms and making them route back to the home is what the marketing division does. This division makes sure the two are wired to the same CRM rather than to two separate versions of the truth.
AI infrastructure is one layer, not three purchases.
The website, the CRM and the dashboards are a single layer with three jobs. The website captures, because it is where an enquiry becomes a record. The CRM remembers, because it holds every interaction with that person from first enquiry to final invoice and beyond. The dashboards tell you what is happening, because a record nobody reads is only storage. Bought separately, by three different people, in three different years, they become three systems that disagree. Built as one layer, they are the rails the rest of the business runs on.
Every bucket reports into this layer. Marketing sends enquiries into it. Operations sends job records, checklists and certificates into it. People sends employee lifecycle events into it, from the signed offer to the completed training to the exit. Governance reads it, because the board pack and the financial oversight are only as good as the records underneath them. Nothing above the rails is asked to keep its own private copy of the truth.
If the layer is missing, each bucket keeps its own truth. Sales knows one number, operations knows another, finance knows a third, and nobody can reconcile them because there was never a shared record to reconcile against. That is the real reason a business of this size often cannot answer a simple question about itself: not laziness, not bad staff, just four honest answers coming from four separate systems. Operations is where most of that record actually gets produced, and governance Structures is where somebody has to own it.
The four buckets
What each one covers, and why it sits where it does.
One. Governance
Entity structures, trusts, family and business constitutions, shareholders' agreements, succession, the board and the non-executive seat, and the financial agents that watch the numbers continuously rather than at month end.
It sits underneath because it is firm infrastructure in Porter's sense: it produces nothing a customer buys, and nothing a customer buys survives without it. Ownership, decision rights and accountability are settled here or they are settled in a dispute later.
Read the foundation
Two. Marketing and sales
The sales machine: AI sales agents that qualify, follow up and book, intelligent call handling, dashboards, and a lighter CRM for clients who do not yet need a full deployment. The marketing division fills the machine, this division builds it.
A primary activity, and the one closest to the customer. It is set apart in the diagram because most of its work routes into the Agentic Digital Marketing division rather than staying here.
Read bucket two
Three. Operations and data
Employee and client apps, checklists that produce a record rather than a tick, digital certificates and job cards, and automation that crosses departmental boundaries so one operational event raises the invoice, updates the pipeline and reaches the financial agents without anyone rekeying it.
A primary activity, and the centre of the whole model. This is also where the value chain audit lives, because the bottleneck holding a business at the same number three years running is almost always here.
Read bucket three
Four. People, the lifecycle
Recruitment, digital onboarding with signed policies and a controlled document vault, training delivered in sequence rather than dumped on day one, and access to labour attorneys for disciplinary and CCMA matters through the HR portal.
Human resource management is a support activity, which is exactly why it cannot be left until last. Every bucket above and below it changes what a person does on a Tuesday morning, and somebody has to hire, train and represent that person.
Read bucket fourWhere financial management and accounting automation sits
It is the question this diagram gets asked most, because finance is not drawn as a bucket of its own. That is deliberate, and here is the reasoning.
- It sits inside bucket one, under governance
- Finance is not a separate activity in this model, it is the evidence layer of the structure. An AI financial management agent runs the same checks daily instead of monthly: reconciling bank movement against the ledger as it happens, flagging a supplier paid twice or a customer over terms, holding today's cash position and covenant headroom against what has been undertaken to a lender, and tracking statutory dates forward with an owner on each one. Put it in its own bucket and it becomes a reporting function. Put it under governance and it becomes the thing that proves the structure is behaving.
- The accounting work itself stays with accountants
- This is not an accounting practice and it is not becoming one. Annual financial statements, tax returns and statutory filings are signed by the people qualified and appointed to sign them. What exists here is a network of accounting practitioners who work with these agents daily and understand structures of this kind, so the agent and the accountant are not fighting each other.
- The operational half sits in bucket three
- An invoice raised from a completed job card, a purchase order matched to a delivery, a job that reaches the financial agents without anyone rekeying it: that plumbing lives in operations and data, because that is where the event actually happens. Finance is where it is checked, not where it originates.
Read what the financial agents do, and what a human still signs.
Sequencing
The order is not a preference. It is the model doing its job.
Governance first, because structure decides who owns what the automation produces. An automated system generates assets the moment it runs: data, records, models, client lists, a documented process. If the trading company holds the client data but the founder personally holds the software subscription, and the family trust holds nothing at all, you have created something valuable that nobody has clear title to. Settle the entities, the shareholders' agreement and the decision rights first, and every later build has an owner from day one.
Operations next, because you cannot automate a process nobody has written down. Automation does not discover how your business works, it executes a description of how your business works. If three technicians each close a job differently and the office reconciles it by memory on a Friday, there is no process to encode, only three habits and a guess. The first half of operations work is therefore writing the process down, and it is common for the written version alone to fix a good part of the problem before a single workflow is built.
Demand after that, because generating enquiries a business cannot fulfil converts a marketing problem into a reputation problem. A campaign that triples enquiries into a business with a two week quoting turnaround does not produce three times the revenue. It produces three times the quotes going out late, the same conversion rate on a worse experience, and a batch of people who now have a first-hand opinion of you they did not have before. Marketing problems are recoverable in a quarter. Reputation in a small South African market takes considerably longer.
People throughout, because every bucket changes what staff do daily. A financial agent flagging exceptions changes what the bookkeeper spends Monday on. A digital job card changes what the technician carries to site. An AI sales agent booking meetings changes what the sales consultant is measured on. If the people work is treated as a final phase, the change lands on staff as a surprise and the honest question is why they should cooperate with it. Brynjolfsson and McAfee (2014) make the case that the durable gains come from machines complementing human work rather than simply standing in for it, and complementarity is not something that happens by itself. It has to be designed into the roles, which means the people bucket runs alongside the other three rather than behind them.
Where automation actually pays
Right first, then up. Never diagonally across the middle.
The sequence above is easier to hold in your head as a picture. Put governance maturity on the horizontal axis, meaning how much of the structure is written down, agreed and actually governed. Put automation intensity on the vertical axis, meaning how much of the work runs without a person pushing it. Every business sits somewhere on that grid, and the four corners behave very differently.
Text equivalent of the matrix. The horizontal axis is governance maturity, meaning structure written down and actually governed, running from low on the left to high on the right. The vertical axis is automation intensity, running from low at the bottom to high at the top. That gives four quadrants.
Bottom left, manual and fragile: it works because one person holds it together, and that person is also the ceiling and the single point of failure. The label reads, most businesses start here.
Top left, fast and wrong: you automated a structure nobody fixed first, so the errors scale too, and nobody can say who owns the data the machine is producing. The label reads, the expensive quadrant.
Bottom right, sound but slow: the structure is right and written down, but it is executed by hand, which is safe and governable and capped by how many hours people have. The label reads, safe and stuck.
Top right, compounding: automation running on a structure that can carry it, where every gain is kept, owned and auditable, and the next one starts from a higher floor. The label reads, where we are taking you.
A thick teal path runs along the bottom from left to right, marked one, fix the structure, and then turns upward along the right hand side, marked two, then automate. The path never crosses the middle diagonally.
Bottom left, manual and fragile. Almost everyone starts here. It works, and it works precisely because one person holds it together in their head: which client is difficult, which job is running late, what was promised on the phone in March. That person is the reason the business functions and also the reason it cannot grow, because they are simultaneously the ceiling and the single point of failure. Nothing is wrong with starting here. Something is wrong with staying here at R60 million of turnover.
Top left, fast and wrong. This is the expensive quadrant. Automation has been laid over a structure nobody fixed, so the errors scale at exactly the same rate as the output. Quotes go out faster and more of them are wrong. Records are produced continuously and nobody can say who owns them, which entity holds them, or which version is authoritative. Say it plainly: this is where most failed automation projects end up. Not because the tools were bad, but because they were pointed at a business that had never agreed with itself how the work is supposed to go.
Bottom right, sound but slow. The structure is right, written down and actually governed, but it is executed by hand. This quadrant is safe, it is auditable, and it is capped: the business can only do as much as the available hours allow, so growth means hiring in proportion to revenue and margin stays where it is. It is a much better place to be stuck than the top left, because the only thing missing is the build.
Top right, compounding. Automation running on a structure that can carry it. Every gain is kept rather than eroded, every asset produced has a clear owner, and the record is auditable, so the next improvement starts from a higher floor rather than from scratch. This is the only quadrant where the work you commissioned last year is still paying you this year.
The point of the diagram is the path, not the corners. You move right along the governance axis before you move up the automation axis. Fix the structure, then automate it. Going diagonally across the middle, tackling both at once because it feels faster, is precisely how businesses land in the expensive quadrant: half a structure, fully automated, with nobody accountable for the difference. Davenport and Ronanki (2018) make a similar observation about organisations reaching for ambitious artificial intelligence projects ahead of the more mundane groundwork, and recommend starting with the unglamorous work instead. The grid is simply that advice drawn as two axes.
The contradictions in this grid are deliberate
Three of the labels pair a good word with a bad one on purpose, because each names a situation that behaves the opposite way to how it sounds. They are worth a moment even if the rest of the page was obvious.
- Fast, and wrong
- Speed is normally the thing you are buying. Here it is the thing hurting you. Automation does not judge the process it is given, it only repeats it, so a flawed process automated becomes a flawed process running thousands of times a day without anyone reading the output. The business that automated a broken structure is now getting things wrong faster than the business that never automated at all, and it paid for the privilege.
- Sound, but slow
- Everything about this position is correct. The structure is right, it is written down, it is governed, and people follow it. That is exactly why it becomes the ceiling: correct work done by hand can only ever go as fast as the available hours, so the better the business gets at doing it manually, the more clearly the hours become the limit. Being right is not the same as being able to grow.
- The expensive quadrant
- The one that costs the most usually returns the least, which is the reverse of what the spending suggests. Businesses land there by treating governance as the boring part to be dealt with later and automation as the exciting part to be started now. The money goes into the visible half, the invisible half never gets fixed, and the resulting system has to be unpicked before anything can be built properly. Doing it in the right order costs less than doing it twice.
- Why the slow thing is what makes you fast
- Governance looks like the brake and behaves like the engine. It is the least urgent-feeling work on the list, and it is what decides whether the automation you buy afterwards keeps paying you or quietly turns into something you have to maintain. That is the whole argument of the grid in one line.
What this does not fix
Automation multiplies. It does not invent.
There are three conditions where the model above will not help, and it is cheaper for everyone if they are said now rather than discovered in month four.
No offer. If the market does not want what you sell, faster delivery of it is not an improvement. Automation raises volume and lowers unit cost, neither of which creates demand that was never there. That is a positioning problem and it gets solved with a decision about what the business is for, not with a workflow.
No margin. If a job loses money at R100 000, it loses money faster when you can process forty of them. Automation compresses cost, it does not manufacture gross profit out of a price that was set too low. Fix the pricing, then automate the delivery, in that order.
No discipline. If agreed processes are optional today, they will be optional inside a system tomorrow, with the added irritation that the system now records exactly who ignored them. Davenport and Ronanki (2018) found that the projects that work in practice tend to be unglamorous process automation rather than ambitious cognitive engagement, and process automation only holds where somebody enforces the process. A tool cannot supply management attention that does not exist.
The common thread is simple enough to state plainly: automation multiplies whatever is already there, including the problems. That is the argument for doing governance first, and it is also the reason this practice will occasionally tell a business to spend the next six months on something we do not sell.
References
- Brynjolfsson, E. and McAfee, A. (2014) The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. New York: W. W. Norton and Company.
- Davenport, T.H. and Ronanki, R. (2018) 'Artificial Intelligence for the Real World', Harvard Business Review, January to February 2018 issue.
- Porter, M.E. (1985) Competitive Advantage: Creating and Sustaining Superior Performance. New York: Free Press.
One clear next step
Find out where your business actually sits on this chain.
A value chain review walks the four buckets in order and reports back on what exists, what does not, and which gap is genuinely holding the business at its current number. You get a written view with the sequence set out, so you can see what has to happen before anything else is worth building. If the answer is that the structure is sound and the real problem is elsewhere, that is what it will say.
Request a value chain reviewOr start at the foundation and read what governance covers first.