the business
WishWash brought wet-cleaning to Egypt in 2019 — no branches, no shop visits, just door-to-door pickup and delivery across New Cairo. Their whole promise rests on one line they put right on the homepage: "We call to confirm every order." Every booking, every price, every exception (leather, carpets, wedding gowns "on inspection") gets a human voice on the phone before a driver ever knocks on a door.
That promise is the brand. It's also the bottleneck.
the night the queue won
It's a Thursday evening in Ramadan, an hour before iftar, and every phone in the WishWash office is ringing at once. Coats are going to the cleaners because the cold snap caught everyone off guard, a jacket-bundle promo just went out on Instagram, and the WhatsApp inbox is scrolling faster than anyone can read it. The site promises a callback "usually within 30 minutes." Tonight it's closer to two hours.
Messages that come in after 9 PM queue for "first thing next morning" — except by morning, the overnight backlog is competing with the day's first wave, and the cycle starts over. When an agent finally does call, the first few minutes go to questions the customer already typed out: which area, which service, which time window. That context lived in a chat thread nobody had time to reopen. And the pricing itself is genuinely intricate — per-item, per-kg, bundles, a whole page of exceptions about what a kg package does and doesn't cover — so a rushed agent occasionally quotes the wrong tier, and untangling it afterward costs more time than the booking itself.
Nothing about the service is broken. The wet-cleaning is still first in Egypt, the drivers still show up, the prices are still fair. What's cracking is the one thing WishWash actually sells: the certainty of "we'll call you back." It's starting to feel like a hold queue.
the call to modernise
WishWash's ops lead reaches out to Modernise, an Anthropic solutions partner, with a narrow ask: don't touch the phone call, fix everything before it. The team spends the first few days just listening — sitting in on the call center, reading a week of WhatsApp transcripts, mapping out exactly where the minutes go. The diagnosis comes back simple: the bottleneck isn't the confirmation call itself, it's everything that has to happen before an agent can make it well.
So the brief that comes out of that week is deliberately narrow: keep the human phone call, the counted items, the price locked in at the door — that's the whole brand. Just make sure nothing sits in a queue untouched, and no agent ever has to reconstruct a conversation or a pricing rule from memory under pressure.
building it
Week 1–2 · Claude Code
An engineer from Modernise pairs directly with Claude Code in the terminal to build the connective tissue: a lightweight service sitting between WhatsApp Business API and WishWash's existing booking system. Rather than guess at pricing, it ingests WishWash's own published rules — per-item, per-kg, bundles, every exception — as a structured knowledge source, so the logic never drifts from what's on the site.
Week 3 · Claude for customer service
With the plumbing in place, Claude takes the first conversation on WhatsApp — any hour, any day, including Friday when the shop is closed. It greets the customer, collects area, service, and preferred window, flags anything priced "on inspection," and gives an instant, accurate price range, all before a human ever picks up the phone.
Week 4 · Pilot
Modernise runs the new flow alongside the old one for a week, watching a handful of real conversations end to end and tuning the handoff — trimming what agents see, adding a missing area to the intake, tightening the "priced on inspection" wording — before switching it on for everyone.
Launch
Every conversation now arrives at the human confirmation call already structured — name, address, service, items, and price expectation sitting in front of the agent before the phone even rings. The call itself hasn't changed. What happens in the five minutes before it has.
Nothing about the brand promise changed. A human still calls to confirm every order, counts every item at the door, and locks in the final price on the spot — that's still WishWash's whole model. What changed is what happens before that call.
the same thursday, months later
Picture the same kind of night — cold snap, a promo running, the inbox filling up fast. A message comes in at 11 PM, long after the office has closed. This time it doesn't wait for morning: the customer gets an immediate reply, their pickup details are already captured, and their price range is already sitting in front of them. When an agent opens the queue the next day, nothing has to be reconstructed. The conversation, the items, the price — it's all there, ready for the one call that still matters most.
what moved
the point
WishWash didn't want a chatbot instead of a phone call — their whole differentiation is that a person confirms everything with you. The win wasn't replacing the human step; it was making sure the human step started from a clean, accurate handoff instead of a backlog.
That's the shape of most good Claude deployments: not "remove the people," but "give the people a better starting point."