Mr Tan built his electrical components distribution business in 1994, working out of a Geylang shophouse with nothing but a fax machine, a Rolodex, and the kind of stubborn determination that defined Singapore's first generation of SME owners. Thirty years later, his company supplies to over 200 B2B clients across Southeast Asia, employs 28 staff, and turns over $8 million annually. By every measure, he succeeded. But when his operations manager resigned last March, taking with her three decades of institutional knowledge stored entirely in her head and a labyrinth of Excel spreadsheets, Mr Tan faced a question he'd been avoiding for years: what happens when the systems that built this business can no longer sustain it?
This is the story of what happened when a 30-year family business deployed its first AI agent. It's not a story about replacing workers or chasing trends. It's about a man who built something real, who feared losing everything that made it special, and who discovered that the next chapter didn't require abandoning the old one.
The Weight of Thirty Years
When we first met Mr Tan at his Ubi office, he was defensive. Arms crossed, jaw set, the posture of a man who'd heard too many pitches from people half his age telling him his business was outdated. "I've been doing this since before you were born," he said within the first five minutes. Fair enough. He had.
His company, which we'll call Tan Electrical Supplies (not the real name, at his request), had survived the Asian Financial Crisis, SARS, the 2008 recession, and COVID. It had outlasted competitors who'd expanded too fast, suppliers who'd cut corners, and economic cycles that crushed businesses with weaker foundations. The formula was simple: know your customers, honour your word, and never let a relationship go cold.
But the formula was also fragile. It depended entirely on people who'd been there long enough to remember that Mr Lim from the Jurong workshop prefers his quotes faxed (yes, still), that the procurement manager at a major construction firm always orders extra cable ties in Q4, and that three specific clients need their invoices sent to different email addresses depending on the project site.
When Institutional Knowledge Walks Out the Door
Mrs Wong, his operations manager of 27 years, didn't leave on bad terms. She retired. Her husband was unwell, her grandchildren needed her, and she'd earned the rest. But when she handed over her responsibilities, the handover document was 14 pages of bullet points that made sense only to her. "Check the blue folder for Keppel orders" — but which blue folder? There were six. "Raymond knows the Changi account" — but Raymond left in 2019.
The first month without Mrs Wong was chaos. Orders slipped. A $45,000 contract was nearly lost because no one knew the client's payment terms had been renegotiated in 2018. The new operations coordinator, a sharp 26-year-old named Rachel, spent more time asking questions than doing her job — not because she was incompetent, but because the answers lived in filing cabinets, old emails, and the memories of people who no longer worked there.
Mr Tan's son, Wei Ming, had been pushing for "digital transformation" for years. Every family dinner became a debate about cloud software, automation, AI. Mr Tan resisted. Not because he didn't understand the technology — he understood enough — but because he didn't trust it. "These systems don't know my customers," he'd say. "They don't know that Mr Goh's father and I played mahjong together for 20 years. You can't automate that."
He was right. You can't automate relationships. But you can automate everything that gets in the way of them.
The Fear Behind the Resistance
Let's be honest about what holds most established business owners back from AI adoption. It's not ignorance. It's not stubbornness. It's fear — and the fear is legitimate.
Mr Tan built his business on a specific kind of value: personal service, deep relationships, the trust that comes from decades of showing up. When someone tells him to "implement AI," what he hears is: everything you built doesn't matter anymore. The world has moved on. You're a dinosaur.
That's not what we told him. What we told him was this: you've spent 30 years building something most businesses never achieve — genuine customer loyalty. But your best people are spending 60% of their time on tasks that have nothing to do with customers. They're chasing invoices, re-keying data, answering the same questions, searching for information that should be at their fingertips. That's not protecting your legacy. That's wasting it.
The Real Question Isn't "Should We Use AI?"
The real question is: what do you want your team doing with their time? Mr Tan's answer was immediate: "I want them talking to customers. Building relationships. Solving problems." Perfect. Then why was Rachel spending four hours a day updating spreadsheets? Why was his sales coordinator, David, manually checking stock levels before every quote? Why was Mr Tan himself staying until 8pm reconciling purchase orders that should have matched automatically?
The resistance to AI often comes from a false choice: either we stay human and personal, or we become automated and cold. But that's not the choice. The choice is between a team drowning in admin work with no time for customers, or a team freed from admin work with more time for customers than ever before. We've written before about why AI should build better jobs, not fewer — and Mr Tan's company became a case study in exactly that principle.
Same team. Bigger output. That's not a slogan. That's what happened.
Starting With the Mess
We don't walk into businesses with pre-built solutions. We walk in with questions. The first two weeks with Tan Electrical Supplies were spent doing something decidedly unglamorous: watching people work.
We sat with Rachel as she processed orders. We shadowed David as he prepared quotes. We asked Mr Tan to walk us through his end-of-month reconciliation process. We documented every spreadsheet, every folder, every workaround that had evolved over 30 years of "we've always done it this way."
What we found was typical for established SMEs:
- Customer information scattered across three systems (an ancient accounting package, Outlook contacts, and a shared Excel file that hadn't been updated since 2021)
- Order processing that required manual data entry in four different places
- A quote generation process that took 45 minutes per quote because David had to check stock, calculate margins, look up customer-specific pricing, and format everything manually
- Invoice follow-ups that happened only when someone remembered — usually after the payment was already late
- Zero visibility into which customers hadn't ordered in months, which products were moving slowly, or which quotes had been sent but never followed up
None of this was Mr Tan's fault. These systems had evolved organically over three decades. They worked — barely — because people like Mrs Wong had memorised the workarounds. But Mrs Wong was gone, and the workarounds were breaking.
The Diagnosis Before the Prescription
We presented our findings to Mr Tan and Wei Ming together. The father-son dynamic was tense — Wei Ming wanted to leap straight to solutions, Mr Tan wanted to defend the status quo. We sided with neither.
Instead, we showed them a simple calculation. Based on our observations, the team was spending approximately 47 hours per week on tasks that added zero value to customers: re-keying data, searching for information, fixing errors caused by manual processes, and chasing payments that should have been automated. At an average fully-loaded cost of $35 per hour, that's $1,645 per week. $85,540 per year. Enough to hire another full-time salesperson — or enough to fund an AI implementation that would give them back those hours permanently.
Mr Tan's response was telling: "But those tasks still need to get done." Exactly. They do. But they don't need to be done by humans. The real cost of manual operations isn't just money — it's opportunity. Every hour Rachel spends updating spreadsheets is an hour she's not spending learning the business, building customer relationships, or developing into the operations leader Mr Tan will eventually need.
Designing the First Agent
We didn't propose a massive transformation. Massive transformations fail. They overwhelm teams, drain budgets, and create resistance that kills projects before they deliver value. Instead, we proposed starting with one agent, solving one problem, and proving the concept before expanding.
The problem we chose: quote generation. It was painful enough to matter, contained enough to manage, and visible enough that success would be obvious to everyone.
Here's what the process looked like before the agent:
- Customer calls or emails requesting a quote
- David checks if the customer exists in the system (sometimes they don't, requiring manual creation)
- David looks up the customer's pricing tier (stored in a separate spreadsheet)
- David checks stock levels in the inventory system
- David calculates margins based on current supplier costs (another spreadsheet)
- David manually creates the quote in Word, formatting it with the customer's details
- David emails the quote and logs it in yet another spreadsheet
- If the customer doesn't respond in a week, David is supposed to follow up (but often forgets)
Average time: 45 minutes. Error rate: approximately 8% (wrong pricing, outdated stock information, formatting mistakes). Follow-up rate: about 60% of quotes never received a follow-up.
What the Agent Actually Does
The AI agent we built — which the team eventually nicknamed "Mrs Wong 2.0," much to the original Mrs Wong's amusement when she heard — integrates with their existing systems and handles the entire quote workflow:
When a quote request comes in (via email or WhatsApp), the agent automatically identifies the customer, retrieves their pricing tier, checks real-time stock levels, calculates margins based on current costs, generates a professionally formatted quote, and sends it — all within 90 seconds. It then schedules automatic follow-ups: a gentle reminder after 3 days, a more direct follow-up after 7 days, and an alert to David if the quote remains unresponded after 14 days.
David's role didn't disappear. It transformed. Instead of spending 45 minutes per quote on mechanical tasks, he now spends 5 minutes reviewing the agent's work, adding personal notes where appropriate, and focusing on quotes that need human judgment — complex specifications, unusual requests, or strategic accounts where the relationship matters more than the efficiency.
If you're curious about how AI agents work in practice, our services page explains the approach — but the short version is this: agents don't replace thinking. They replace typing.
The Moment Everything Changed
Three weeks after deployment, something happened that shifted Mr Tan's entire perspective.
A long-standing customer — a mid-sized contractor Mr Tan had known for 15 years — called to place an urgent order. They needed specific cable specifications delivered to a Tuas site by the next morning. In the old days, this would have triggered a scramble: checking stock, calling suppliers, manually creating the order, coordinating with the warehouse, hoping nothing fell through the cracks.
Instead, David pulled up the customer's profile in the new system. The AI agent had already flagged that this customer's typical order pattern suggested they'd need restocking soon. Stock was available. Pricing was pre-calculated. The quote was generated in under a minute. The order was confirmed, the warehouse was notified automatically, and delivery was scheduled — all before Mr Tan finished his coffee.
The customer called back an hour later. Not to complain. To ask how they'd become so fast. "You used to take half a day for quotes," he said. "Now you're faster than the big guys. What happened?"
Mr Tan didn't say "AI." He said, "We've been working on our systems." Which was true. And which was exactly the right answer. Customers don't care about your technology. They care about their experience.
The Numbers After 90 Days
We measure everything. Not because we're obsessed with metrics, but because established business owners like Mr Tan need proof, not promises. Here's what the data showed after three months:
- Quote generation time: reduced from 45 minutes to 4 minutes (including human review)
- Quote error rate: reduced from 8% to under 1%
- Follow-up rate: increased from 60% to 100% (automated follow-ups never forget)
- Quote-to-order conversion: increased from 34% to 41% (faster quotes close more deals)
- Time saved per week: 12 hours on quote-related tasks alone
- Customer complaints about slow response: zero in the past 60 days (previously 2-3 per month)
But the number that mattered most to Mr Tan wasn't on any dashboard. It was this: David, his sales coordinator, had started visiting customers again. With 12 fewer hours per week spent on quotes, he had time to do what salespeople should do — build relationships. In the first quarter after deployment, David brought in three new accounts. Not because of AI. Because AI gave him back the time to do his actual job.
Expanding Without Overwhelming
Success with one agent created appetite for more. But we were careful. The worst thing you can do with an established business is move too fast. These aren't startups with nothing to lose. They have customers, reputations, and staff who've been doing things a certain way for years. Change needs to be digestible.
The second agent we deployed handled accounts receivable follow-up. Before, invoice chasing was Rachel's least favourite task — awkward calls to customers asking for money, easy to postpone, impossible to systematise. The agent now sends automated payment reminders at 7, 14, and 21 days overdue, escalating tone appropriately. It flags accounts that need human intervention (disputes, unusual delays, strategic relationships where a call from Mr Tan himself might be appropriate).
Result: average days sales outstanding (DSO) dropped from 52 days to 38 days. Cash flow improved. Rachel stopped dreading Mondays.
The third agent — deployed six months in — was more ambitious. We built what we call an "AI Brain" for customer intelligence. It synthesises data from across the business to answer questions like: which customers haven't ordered in 90 days? Which products are trending up or down? Which customers might be at risk of churning based on order pattern changes? This information had always existed in the business — scattered across systems, visible only to people who'd been there long enough to recognise patterns. Now it's surfaced automatically, actionably, in real time.
Wei Ming, the son, now runs a weekly review meeting using the AI Brain's insights. It's the first meeting in the company's history that's driven by data rather than gut feel. Mr Tan attends. He doesn't always agree with the data. But he's learning to ask better questions.
What the Team Actually Thinks
We interviewed each team member six months after deployment. Their responses matter more than any metric.
Rachel (Operations Coordinator): "I was scared at first. I thought they were bringing in AI to replace me. But it's the opposite. I used to feel like a data entry clerk. Now I feel like I'm actually managing operations. The boring stuff happens automatically. I spend my time on problems that need thinking."
David (Sales Coordinator): "The quote agent is like having a really fast assistant who never makes mistakes. I still review everything — I'm not just pressing approve blindly. But I'm reviewing, not creating. It's a completely different job. Better job."
Mr Tan (Founder): "I was wrong. I thought AI meant becoming like those big corporations — cold, impersonal, everything automated. But my customers don't know we're using AI. They just know we're faster. And my staff are happier. That's what matters."
Wei Ming (Son, now Operations Director): "Dad and I fought about this for years. But watching him come around — seeing him actually excited about the business again — that's been the best part. He's not just maintaining anymore. He's building again."
Same team. Bigger output. It's not just a phrase. It's what happened.
The Conversation About Succession
Here's something we didn't expect. Six months into the AI implementation, Mr Tan and Wei Ming had a conversation they'd been avoiding for a decade: succession planning.
For years, Mr Tan had resisted handing over control because he didn't trust that Wei Ming understood the business deeply enough. And honestly? He was right. Wei Ming had ideas — good ideas — but the institutional knowledge, the customer relationships, the operational nuances — those lived in Mr Tan's head and Mrs Wong's spreadsheets.
The AI Brain changed that. For the first time, the business's collective knowledge was documented, searchable, and transferable. Wei Ming could see which customers were most valuable, which products had the best margins, which patterns indicated problems. He wasn't guessing anymore. He was learning — faster than he ever could have by osmosis alone.
Mr Tan set a date: he'll step back to an advisory role in 18 months. Not because the AI made him redundant — he's more valuable than ever for relationships and strategic judgment — but because the AI made succession possible. The business is no longer dependent on any single person's memory.
That's legacy preservation. Not replacing the founder. Protecting what the founder built.
What This Means for Your Business
If you've built something over 20 or 30 years, you've earned the right to be skeptical of consultants promising transformation. You've seen trends come and go. You've watched competitors chase shiny objects and fail. Your caution is an asset, not a weakness.
But here's what we've learned working with established businesses across Singapore: the question isn't whether to adopt AI. The question is whether your current systems can sustain another 10 years. Can your business survive when your longest-serving employees retire? Can you compete when younger competitors respond to customers in minutes while you take hours? Can you grow when your best people are buried in admin work?
If you're considering taking the first step, the Enterprise Development Grant can fund up to 50% of qualifying transformation projects. We've helped multiple SMEs structure their implementations to qualify. It doesn't make the decision for you — but it makes the decision easier.
The Principles That Made This Work
Mr Tan's transformation succeeded because we followed principles that matter for any established business:
- Start small, prove value, then expand. One agent. One problem. Visible results. Then build from there.
- Respect what exists. Thirty years of business relationships aren't inefficiencies to be eliminated. They're assets to be protected and amplified.
- Involve the team from day one. Rachel and David weren't told about the AI. They were consulted, trained, and given ownership. Their buy-in made implementation smooth.
- Measure what matters to the owner. Mr Tan didn't care about "digital transformation metrics." He cared about customer relationships and staff wellbeing. We measured those.
- Move at the business's pace, not the technology's. We could have deployed faster. We chose not to. Sustainable change beats impressive demos.
As we've discussed in our analysis of why most AI projects fail in year one, the technology is rarely the problem. The implementation approach is. Established businesses need partners who understand that their history is a feature, not a bug.
The Shophouse Still Stands
Mr Tan sold the original Geylang shophouse in 2008. But he kept a photo of it on his office wall. When we finished our six-month engagement, he pointed to it during our final meeting.
"My father thought I was crazy to start this business," he said. "He wanted me to be a teacher. Stable job. Pension. But I wanted to build something. Thirty years later, I almost let it fall apart because I was too proud to admit the world had changed."
He paused. "The shophouse is gone. But what I built there — the relationships, the reputation, the way we treat customers — that's still here. The AI didn't replace any of that. It just... made room for more of it."
That's the point. AI doesn't erase legacy. It extends it. Your best people shouldn't be chasing invoices. They should be doing what made your business special in the first place.
If you've spent decades building something real, and you're wondering whether there's a way forward that doesn't require abandoning everything that got you here — there is. It starts with a conversation, not a sales pitch. Talk to us at abuzz.sg/contact. We'll listen first. That's how this works.
