You can’t automate a process you haven’t structured yet
|
|
Plus: The three questions your pilot never asked (but production will)
|
|
|
|
|
From the aibl team
|
|
The standard AI playbook is to start simple: pick a well-known process, automate it, prove the value, then expand.
Josh Clement-Sutcliffe is AI and Automation Product Manager at Zencargo, a digital freight forwarder. A single shipment can involve dozens of parties across multiple countries, each with their own systems, regulatory requirements and document standards.
Zencargo started down the happy AI path but the variables were too many and the context too fluid. Communications still run on email, fax and phone, so when something happens like a Suez Canal blockage, every route and every customer commitment has to be renegotiated overnight. That’s not in the playbook.
So rather than finding a new tool or a better model, they focused on getting the inputs clean and structured enough for AI to act on reliably.
The standard approach works if the underlying process is stable enough to isolate, but that’s not the case in a lot of mid-market businesses. The real task becomes building enough structure into fragmented operations that AI can do more than assist at the edges. Josh’s account is below.
|
|
When AI meets operational chaos, the happy path is off a cliff
|
|
Josh Clement-Sutcliffe is AI and Automation Product Manager at Zencargo, a London-based digital freight forwarder
The standard advice is to pick a simple process, automate it, scale from there. Global freight forwarding doesn’t have a simple process. A single shipment can involve dozens of parties across multiple countries, each with their own regulatory requirements, systems, and standards. Communications still run on email, fax and phone. There’s no shared infrastructure or common data format, and nothing that makes one supplier’s documentation look like another’s. So there’s no playbook to reach for. It’s a business where a container ship can block the Suez Canal, and every route, timeline and customer commitment has to be renegotiated overnight.
Zencargo’s early automation work targeted the happy path anyway, where everything arrives in the right format, from the right party, through the right channel. But automating that first, then expanding outward didn’t work. “The AI and automation just couldn’t figure out the context and what to do given the amount of variables which had to go into those particular processes.”
Josh describes falling into the sunk cost fallacy on a couple of projects, which is natural enough when there’s no industry precedent. The evidence eventually pointed one way: go back to the data.
Last August, they put a plan in place. First, data foundations: getting inputs clean enough for AI to act on. Second, visibility and trust: letting operators see what the system is doing. Third, what he calls partnership, where AI automates predictions, risks and optimisations at scale. 2025 was the foundational year. 2026 is about operationalising and scaling, with visibility and trust as the focus.
|
|
|
|
Watch the full video interview:
|
|
|
|
AI in practice
|
|
|
What broke when an agent moved from pilot to production
A mid-market professional services firm moved an AI agent into their live billing support queue earlier this year. The agent had one job: triage tickets, draft first responses, and pull context from previous interactions so the team wasn’t starting cold every time.
After a successful eight-week pilot on a controlled set of tickets, they connected it to the live queue. What followed wasn’t a disaster – no client received something they shouldn’t have – but the failure was subtler than that, and harder to fix.
What the pilot hadn’t tested
Within a few weeks, three things were missing.
The first was an audit trail. The agent was updating records and moving tickets, but sometimes it wasn’t clear what it had done, or why. When a team lead tried to reconstruct what had happened to one billing query, two people spent the better part of an afternoon piecing it together from different messages and system timestamps.
The second was how to review it. During the pilot, three people handled review informally and it worked. In production, eight people were touching the queue. Some read every draft carefully. Others waved through responses with minimal checks, because it seemed to be working and they had other things to do. One message went out that was factually accurate but struck the wrong tone with a client chasing an overdue invoice.
The third was measurement. When the ops director asked what the agent had actually delivered, nobody could point to a KPI that had moved. The agent was running, but nothing about it could survive a budget conversation.
The agent wasn’t switched off, but it drifted into limbo. The team went back to their old workflows, occasionally checking the agent’s output after the fact, if at all.
We see this pattern often at aibl. Pilots rarely test everything around the agent: the audit trail, the review process, the ability to show what it actually did.
|
|
|
|
|
Product spotlight of the week
|
|
|
We covered Salesforce back in November around their Einstein GPT feature set, and this week they caught our attention again with Agentforce.
The pitch for mid-market teams is that AI agents now sit across the whole platform – handling lead qualification, pipeline follow-up, customer queries and routine service tasks without a human in the loop. Salesforce has closed over 18,500 Agentforce deals since its launch in late 2024, with UK adoption accelerating.
The problem it solves is one most mid-market sales and service teams know well: plenty of CRM data, not enough headcount to act on it.
Starter Suite runs at £20 per user per month, with Pro Suite at £80. Implementation scales with what you need. Worth talking to the Salesforce team about where to start.
|
|
|
|
Copyright (C) 2026 aibl. All rights reserved.
|
|
