In the first piece in this series, we said venture backability is a two-part test. You have to be able to lead a clearly defined, fast-growing niche, and you have to stay the leader once you are winning. That piece was about leading. This one is about lasting: building a startup moat.

It matters more now than it did three years ago. When building software was expensive, a good product was itself a form of protection. That is no longer true. So “we have built great features” has stopped being evidence that a company is backable.

This piece asks three questions:

Here is the part that is easy to miss. For three decades, writing software was slow and expensive. It took scarce, well-paid engineers and long build cycles, which meant real time and real money.

That expense quietly did a second job. It kept competitors out.

To copy a serious product, a rival had to spend the same years and the same millions. The difficulty of building was itself a wall around the business. The whole software-as-a-service model leaned on that wall: build it once at great expense, sell it over and over as a subscription, and let the cost of rebuilding hold rivals back and keep margins high.

That wall is gone. Shipping is faster, and cloning is cheaper, so a great product is now table stakes rather than a defence. The interface is not the moat. The workflow is.

Founders sometimes hear this and assume it applies to someone else. It does not, and the reason is uncomfortable: fast growth is the loudest possible signal that there is real money in your category.

It attracts three kinds of competitors.

Copycats, who simply replicate what is working. Incumbents, who cannot ignore a fast-growing pocket of profit sitting inside their own buyer base. And adjacents, who already sell to your customer and can add your feature set as a bundle.

So in a market where software is cheap to build and getting cheaper, the default outcome of traction is attention, and attention becomes competition. That is not a risk you can avoid by executing well. It is what executing well produces.

This is why durability is a venture question and not a product question. A company can grow early without any protection at all. But if nothing pins customers to you, growth simply invites copycats and incumbents to compete your margins away. Building fast is not a right to win.

The first is build-cost compression. If your edge was that you could do something cheaper than the incumbent, AI removes that advantage, and it removes it for everyone at once.

The labour-arbitrage model of the last decade is the clearest global illustration. Those businesses worked because the unit being sold was human output: rent a strong engineer in a lower-cost market, sell that capacity into a higher-cost one, and keep the spread. AI changes the unit. When one strong engineer with AI tools ships what used to take a small team, the buyer’s cost per feature falls everywhere, and the spread that powered the model shrinks fast.

We have watched the local version of this. An integration-software business we sat with had a simple early hook: enterprise-grade integration at a price the incumbents could not serve. As AI made integration easier to assemble, that price advantage narrowed, and some customers’ own technical teams built the integrations themselves. The company now has to rebuild its protection around a specific workflow, customer data and distribution, rather than around cheaper engineering.

The second way is more structural: the platform becomes the product. The AI provider does not politely stay underneath you as neutral infrastructure. It can move up into the customer’s workflow and sell the finished product itself. That risk is highest where the output is mostly text, code or data, and where switching tools is easy.

Legal software is the cleanest current example. In May 2026, Anthropic shipped Claude for Legal: twelve practice-area plugins and more than twenty connectors into the systems legal teams already run on. That is not a general chatbot with a legal prompt attached. It is a workflow product, and it changes the competitive set for every legal AI startup.

Look closely at what that does to a company like Harvey. Harvey is a named integration partner in that launch and a competitor to it at the same time. That is the position to understand, because it is the position most application-layer companies will end up in: you build on the platform, you distribute through the platform, and the platform can bundle your core use case into its base product and sell it to the same buyer.

When the underlying model is similar for everybody, a standalone application survives only if it owns something the model provider cannot quickly copy.

Protection no longer lives in the feature list. It lives in the parts of the business that customers, new entrants, incumbents and AI providers cannot quickly copy. In practice, it comes from six places.

Customer-specific data. The product gets better because it learns from that customer’s actual history, labels, decisions and usage. A generic model does not have it, and a new competitor cannot buy it. Test: if a competitor cloned your interface tomorrow, would they still be worse because they lack the data your product has learned from?

Being embedded in daily work. The product becomes part of how work gets done: approvals, integrations, routines, reports, handovers. Replacing it means changing behaviour and retraining people. Test: what breaks, slows down or becomes riskier if you are switched off for a week?

Owning the operating memory. The product becomes the place where the important work and records live. Leaving means moving history, context, permissions and audit trails, not exporting a file. Test: where does the truth live today, and how painful would it be to move it?

Access and trust. AI copies features faster than it copies a trusted route into a specific buyer base. Test: do you have a repeatable channel a competitor cannot cheaply access, through partners, references or procurement relationships?

Local and implementation depth. Global platforms solve the larger, more generic problems first. The harder work here is usually compliance, reporting, local rails, security review, legacy integration, and making the product work safely in the buyer’s real operating context. Test: what must be true before a South African buyer can say yes, and how long would a global product take to catch up?

Together, these create switching costs that actually hold. The customer stays not because leaving is impossible, but because leaving would be risky, disruptive, expensive or slow.

The sixth is different in kind. Doing what an incumbent cannot copy without hurting itself. Sometimes the advantage is that your product is simpler, cheaper or faster to adopt, and sold in a way the incumbent cannot match. If you sell a transparent, low-cost product that a customer can adopt in an afternoon, an incumbent who earns its margin on complexity, bundles and long implementation projects cannot copy you without breaking its own economics. Test: what would the incumbent have to damage, in pricing, channel or margin, to copy you?

There is a local advantage here worth naming, and it is the opposite of the one founders usually claim.

Many South African business buyers, particularly in regulated or operationally complex sectors, run on legacy systems, long procurement cycles and relationship-based trust. A new entrant has to clear compliance and security review, integrate into entrenched systems, and earn trust before anyone can say yes.

That friction is usually described as what makes selling here hard. It is also what makes leaving hard. Once you are properly embedded, the same friction that slowed you down becomes part of your protection, but only if you design for it deliberately rather than treating it as an obstacle to be minimised.

So the test is simple. If a global incumbent launched here tomorrow at a low price, what would keep your customers from switching? If the honest answer is integrations, compliance posture, implementation depth and trust, you are building the kind of stickiness this market rewards.

Protection does not stay strong on its own. Patents rarely save a startup, and any advantage gets weaker once competitors can see what is working. The only durable defence is to keep making the company harder to replace, faster than others can copy the surface.

Make it operational. Every quarter, ship at least one thing that makes existing customers more embedded, and one thing that makes new customers easier to reach.

On the embedding side, that looks like a deeper integration into the customer’s system of record, a compliance or security approval that becomes reusable, workflow automation that removes manual work, or data capture that makes the product smarter over time. On the distribution side, it looks like a partner channel that produces repeatable leads, a reference network inside a specific buyer community, a wedge into enterprise procurement, or a packaging move an incumbent cannot match without breaking its own model.

The question to hold yourself to: if a well-funded competitor copied your interface in sixty days, what advantage would still be getting stronger six months later?

If a customer left tomorrow, what would they lose besides features?

Data: what history, labels, context and decision memory do they lose? Daily work: what process breaks, slows down or becomes riskier? Records: where does the truth live today, and how painful would it be to move it? Trust and access: what relationships, approvals and credibility would a replacement have to rebuild? Local depth: what would a replacement have to clear again before a South African buyer could say yes? Add those up, and you have the real switching cost: how long leaving would actually take, once training, migration, compliance and internal politics are counted. Then one question for the rival rather than the customer: what would an incumbent have to damage, in pricing, channel or margin, to copy you?

If the honest answer is “not much,” you have a head start rather than a moat. In this era, head starts are measured in weeks.

That is the bar. Before you raise, answer the question that decides everything else now: if someone copied your product tomorrow, why would customers still choose you? Clear it, and the two-part test from the first piece is fully passed. You can lead, and you can last. You are backable.

Which changes the question again. If you are going to raise, whose money actually fits your company? That is the next piece.

If you are a seed-stage B2B venture raising funding, connect with Sizwe and the team at 3 Capital Ventures.