AutoSpark finds the one place AI or automation is genuinely worth applying in an established business, then builds and deploys it. Led by Patrick Nesbitt, CA(SA), CFA and a former private-equity investor.
AutoSpark works with established small and mid-sized businesses that suspect AI could help but cannot say where. Most of them have been pitched a platform, a copilot or a transformation programme, and none of those start by asking what the business actually loses money on.
The work starts with interviews. We talk to the people doing the job, not only the people who own the process on paper, and we look for the places where work repeatedly gets stuck, gets rekeyed, or waits on someone. Those places are then ranked by what they plausibly cost per year. Only the ones where the arithmetic shows a clear payback get built.
AutoSpark is led by Patrick Nesbitt. He qualified as a CA(SA) in 2012 and became a CFA charterholder in 2017. He audited banks and investment firms at KPMG, covered listed companies in equity research at Macquarie, and then moved to mid-market private equity, where he worked on more than 40 transactions and chaired an alternative investment manager.
That background is the reason the method leads with cost rather than capability. An automation that saves four hours a month in a place nobody was waiting is not worth building, and an investor is trained to notice the difference. AI is treated here as a capital allocation question, which is to say it has to clear a hurdle rate like anything else.
Surface is for advisors investigating how an operation works before recommending a change. Its product site explains the discovery approach and shows an illustrative engagement. Use that walkthrough to examine the process, finding and source account together; the fictional company and illustrative effort model are not customer results.
Start with a specific decision, the people who can describe the work, and the records that could confirm or challenge their accounts. In the sample handoff, sales considers the brief sent while delivery still has to reconstruct changes from emails. That difference suggests a follow-up question, not proof that more staff or new software is needed. A reviewer still has to check the records, assess alternative explanations and decide what to change.
A worksheet may be enough when the task is bounded and one owner can maintain the evidence. A structured investigation becomes useful when accounts conflict, responsibilities cross teams, or the recommendation depends on evidence that has not been gathered. A formal audit, legal opinion or specialist engineering assessment needs the appropriate discipline and scope.
The available next step from our resources is to discuss the investigation. Send the question and the gap you need to resolve; scope and access are agreed through that conversation. The resource enquiry does not create a Surface account or upload a completed worksheet.
It is not an AI audit, and it is not a transformation programme. There is no licence to buy and no platform to adopt before anything useful happens. If the interviews show that the honest answer is that nothing here is worth automating yet, that is a legitimate result and it gets said plainly.