--- title: Process automation beats AI hype. SiliconANGLE said it four times this week. description: "SiliconANGLE ran four pieces between April 30 and May 4 saying the same thing in slightly different words. The enterprise value isn't in flashier models. It's in process work that ships, holds up under load, and connects to the systems that already run the business. After 120+ AI projects across 15+ industries, I keep seeing the same pattern in UK SMBs. The teams winning at AI didn't pick the cleverest tool. They picked the boring process first. This is the four-step \"boring first\" method, with one client example." canonical: https://richardbatt.com/blog/process-automation-beats-ai-hype-2026 date: 2026-05-05 author: Richard Batt tags: [Process Automation, AI Strategy, SMB AI, Reliability] type: blog_post --- # Process automation beats AI hype. SiliconANGLE said it four times this week. _SiliconANGLE ran four pieces between April 30 and May 4 saying the same thing in slightly different words. The enterprise value isn't in flashier models. It's in process work that ships, holds up under load, and connects to the systems that already run the business. After 120+ AI projects across 15+ industries, I keep seeing the same pattern in UK SMBs. The teams winning at AI didn't pick the cleverest tool. They picked the boring process first. This is the four-step "boring first" method, with one client example._ **Richard Batt** — AI implementation specialist. 120+ projects across 15+ industries, serving SMBs (5-200 employees) worldwide from Middlesbrough, UK (working globally). Contact: richard@richardbatt.com · https://richardbatt.com If you only read one trade publication this week, the message from SiliconANGLE was hard to miss. Across four separate pieces published between April 30 and May 4, the recurring takeaway was the same: enterprise AI value comes from process work, not from the flashier model. Reliability beats capability. Boring beats novel. The vendors quietly making money are the ones that connect AI to existing systems, not the ones unveiling another foundation model. That message lands differently when you're a UK SMB owner trying to decide what to do this quarter. The enterprise lesson translates almost perfectly to small businesses: pick the boring process before you pick the model. The teams winning at AI in the SMBs I work with didn't shop for tools. They picked one specific workflow, mapped it carefully, and wired in AI as the dull middle step. That's the entire pattern. After 120+ AI projects across 15+ industries, I've watched dozens of UK SMBs try the opposite approach. They buy ChatGPT Teams, then subscribe to Microsoft Copilot, then pilot a chatbot from a different vendor on top. Six months later, nothing has changed in how the work actually gets done. The pattern is so consistent it's almost a control group for the SiliconANGLE thesis. ## What SiliconANGLE actually said Across four pieces in five days, SiliconANGLE's coverage hit the same notes. Process automation, not AI capability, is where enterprise value is being created in 2026. The pieces covered different angles but landed in the same place. The April 30 piece argued that enterprise process automation is reliability over capability, framing AI agents as good only when they sit inside well-defined process boundaries. The May 4 pieces (two of them, on different topics) tracked enterprise buyers prioritising tools that connect AI to existing workflows over tools that demonstrate novel reasoning. The April 29 article on agent infrastructure made the same point from the vendor side: the companies winning agent budget are the ones doing process integration, not the ones with the cleverest models. That's a useful pattern. When four different journalists at one publication independently land on the same theme inside a week, it usually means the field has shifted and the early data is now visible. For SMB owners, the early data is what matters. ## Why "boring first" works at SMB scale The reason process work beats AI hype at SMB scale is that small businesses don't need bleeding-edge model capability. They need a workflow that runs every Tuesday morning without somebody noticing it ran. The capability ceiling of GPT-4-class models is already three layers above what most SMB workflows actually require. So adding more capability doesn't change anything. Adding more reliability changes everything. A 30-person services firm I worked with last month had a quote-generation workflow that took 9 hours of senior associate time a week. The senior associate was using ChatGPT for parts of it but the workflow lived entirely in his head. Each quote required him to paste a client brief into ChatGPT and draft a response. He then moved the output through Word for formatting and the CRM for tracking before sending. The AI was already doing a meaningful portion of the cognitive work. The hours weren't being saved because the process wrapped around the AI was still manual. So we redesigned the process. The CRM now sends the brief to a Claude prompt automatically. The prompt drafts the quote against a structured template. The output flows into a templated Word document and into the CRM as a draft email. The senior associate reviews and sends. Total weekly time on the workflow dropped from 9 hours to 90 minutes. Net saving per week: 7.5 hours. Notice the model didn't get cleverer. The process around the model got tighter. That's the SiliconANGLE point applied to SMB scale. ## The four-step "boring first" method Here's the four-step method I now run on every SMB engagement. It came out of the pattern across the last 50 client projects and it maps cleanly onto the SiliconANGLE thesis. **Step 1: pick the workflow that costs the most hours.** Not the workflow that's the most exciting. Not the workflow your sales team thinks would impress clients. Pick the one that consumes the most senior team-member time per week. In SMBs that's almost always one of: quote generation, proposal writing, customer follow-up, invoice processing, or report compilation. **Step 2: map the manual version end to end.** Write down every step. The handover points. The systems involved. The places where the work stalls because someone is waiting for an email reply. This is where most SMBs skip ahead and lose. The map is what tells you where AI fits, and where AI doesn't help because the bottleneck is actually somewhere else. **Step 3: identify the dullest middle steps.** AI does its best work on the steps that are routine and structured and repeated often enough to learn from. Drafting from a brief. Summarising a document. Categorising an inbound message. Pulling specific fields out of unstructured text. Those are the steps to give to a model. The judgement calls and the relationship work stay with the human. **Step 4: connect the AI to the systems that already exist.** This is where most pilots die. The AI works fine in isolation. It doesn't get used because pasting things in and out of it costs more time than it saves. Use Zapier, Make, or n8n to wire the AI into the CRM, the email system, the document store. The connection is the work. The model is the easy part. That's the entire method. Four steps. Total elapsed time on a typical SMB project: 8 to 14 working days. Total cost: usually £2,000 to £6,000 if you bring in someone like me. Lower if you have a technically curious operations manager and a few weekends. ## Where this goes wrong I should be honest about where the "boring first" method fails. There are two failure modes I see often enough to flag. The first is when the process being automated isn't actually broken in the way you think. Sometimes the 9 hours a week being eaten by a workflow is because the brief format is wrong, or because two systems aren't talking to each other, or because the team hasn't been trained on the existing CRM. AI on top of those problems doesn't help. It paves over the real issue and the savings don't materialise. The second is when the manager who runs the team using the workflow is sceptical of AI. Gallup's 2026 data is clear: teams whose managers actively support AI are 98.7 times more likely to say AI has changed how they work. Without that manager support, the workflow gets quietly reverted within 90 days. The fix is to do the AI conversation with the line manager first, not the founder or the IT person. If you can avoid those two failure modes, the four-step method ships. It's not glamorous. It doesn't make for an exciting LinkedIn post. But it's the pattern that actually saves hours, week after week. ## What this means for your week If the SiliconANGLE thesis is right, and the early SMB data I see strongly suggests it is, the move this week isn't to buy a new AI tool. The move is to identify your highest-cost manual workflow and start mapping it. Twenty minutes with a senior team member and a whiteboard usually surfaces the candidate. Half a day of process mapping usually surfaces where AI fits. The rest is plumbing. Most SMBs lose this race because they spend three months researching tools and zero months mapping processes. The few that win it spend three weeks mapping, three weeks building, and ship. If you want a structured way to find your one workflow worth redesigning, the AI Roadmap audit is the fastest path. It's at https://richardbatt.co.uk/roadmap. ## Frequently asked questions **What's the difference between AI automation and process automation?** Process automation is the broader category: tools and systems that move work through a series of steps without human intervention. AI automation is a subset where one or more steps in that process is handled by an AI model. The SiliconANGLE point is that the value sits in the process design, not in whether any particular step happens to use AI. For most SMBs, the right mix is automation tools (Zapier, Make, n8n) doing the connections, with AI handling the steps that need language understanding. **Why do most AI pilots fail in UK SMBs?** The pattern I see most often is tool-shopping before process-mapping. SMBs buy ChatGPT Teams, Copilot, or a vendor product and try to find places to use it. That backwards-fits the model to the work and almost never produces durable savings. The "boring first" method runs the other direction: pick the workflow, map it, then choose the tool. Across the projects I've shipped, that ordering produces 5 to 12 hours a week of saved time per workflow, while the reverse ordering produces a tool licence nobody uses. **Do I need a developer to wire AI into my existing systems?** For most SMB workflows, no. Zapier, Make, and n8n all let you connect AI models to common systems (Microsoft 365, Google Workspace, HubSpot, Salesforce, Xero, Sage) without writing code. The work is in mapping the steps and writing prompts that produce reliable output. A technically curious operations manager can do most of it. Where you do need a developer is when the AI has to integrate with custom or legacy systems, or when the volume is high enough to justify a production-grade pipeline. --- ## More about Richard Batt Richard Batt is an AI implementation specialist who helps businesses deploy working AI automation in days, not months. 120+ projects across 15+ industries. ### Key pages - [Home](https://richardbatt.com/) - [About Richard](https://richardbatt.com/about) - [Blog](https://richardbatt.com/blog) - [Contact](https://richardbatt.com/contact) - [Subscribe](https://richardbatt.com/subscribe) ### Contact - Email: richard@richardbatt.com - Location: Middlesbrough, UK (working globally) - Website: https://richardbatt.com