Eight months into actively working with the Fund I portfolio, we have enough cross-company visibility to start identifying the workflow automation patterns that appear repeatedly — not as coincidences but as structural features of how AI-native software embeds in enterprise operations. I want to be careful here: these are observations from a small portfolio over a short time period, not industry research. But the patterns are sharp enough to be worth writing down, both because they inform how we think about new investments and because they are genuinely useful for founders building in adjacent spaces.
The first pattern is what we call the "human-in-loop anchor." Every one of our portfolio companies is automating a previously manual process, but none of them are selling full automation. They are selling automation with a configurable human review layer. Workist processes purchase orders automatically but surfaces low-confidence extractions for human approval. Candis categorises transactions automatically but keeps the tax adviser in the loop for ambiguous entries. This is not a product limitation — it is a deliberate design choice driven by enterprise buyers' unwillingness to remove human accountability from business-critical processes entirely. The acceptance threshold for "the machine made this decision autonomously" is much lower in German enterprise than in US tech. The products that are gaining traction have built the human-in-loop as a first-class architectural feature, not as a concession to risk-averse buyers.
The second pattern is integration as the primary adoption barrier. The technical capability of the automation product is almost never the reason for slow adoption. The reason is always the integration surface: connecting to an SAP system that was customised ten years ago and is not well-documented, authenticating against an on-premise identity management system that predates OAuth, extracting data from a legacy document management system that has no published API. Every portfolio company has a war story about a deal that stalled for three months because of an integration problem that should have taken a week. The implication for product strategy is that integration infrastructure is not a growth feature — it is a sales prerequisite, and it needs to be resourced accordingly at seed stage even when it feels like distraction from core product development.
The third pattern is the champion-to-committee dynamic in enterprise sales. Deals typically begin with a single internal champion — an operations director who has been trying to automate a specific process for two years and finally found a product capable of doing it. But before the contract can be signed, the champion needs to bring in IT security for a data processing assessment, finance for budget approval, and often HR or the works council if the automation touches any employee-facing workflow. The founder's job is not just to convince the champion but to arm the champion with the materials they need to sell internally. This is a selling motion that most early-stage B2B SaaS teams underinvest in, particularly technical founding teams who focus disproportionately on the product capability rather than the sales process around it.
The fourth pattern, which cuts across all three of the above: the companies in our portfolio that are growing fastest are those where the founding team has direct personal experience with the process they are automating. Not theoretical knowledge, not market research — actual experience having worked inside the function they are now selling to. This is partly because product design is better when it is informed by lived experience rather than user interviews alone. But more importantly, it is because the first customers are often found through the founding team's personal network in the relevant function, and those first customers are willing to invest in onboarding a new product because they trust the founders' understanding of their problem. That trust cannot be manufactured through sales process — it is a byproduct of genuine domain depth.