AI Automated Solutions Zimbabwe

Customer support automation

Instant answers for the twenty questions, a person with context for everything else.

Support is tiered or it is chaos. Tier 0 answers from your documents in seconds. Tier 1 looks things up. Tier 2 is a named person who receives the transcript and the customer's history instead of a cold 'hi'.

The Zimbabwean situation

What usually goes wrong

"Are you open?", "How much is…?", "Where is my order?", "Do you deliver to Chitungwiza?", "Can I pay in ZiG?". The same twenty questions, on WhatsApp, all day, answered by whoever is free, differently each time. The real complaint arrives at 18:30 and is read at 08:00 by someone who does not know the history.

Staff churn means the answers live in people, not documents. Price changes with the exchange rate mean last week's answer is wrong. And under the Consumer Protection Act, a mishandled complaint is a legal matter, not a bad review.

What we build

What gets built

  • Tier 0 — instant answers. Hours, locations, delivery areas, payment options, prices from the current list, policies. Drawn from documents you own and can edit.
  • Tier 1 — lookups. Order status, delivery ETA, account balance, booking details, from the systems that hold them.
  • Tier 2 — people. Complaints, refunds, anything the flow is unsure about, and anyone who asks for a human. Routed with transcript and history, within hours you set.
  • Roster and hours. Who is on, when; after-hours acknowledgement with a next-morning promise that is kept.
  • Improvement loop. Weekly list of unanswered questions; monthly handled/handover/first-reply metrics.

How it works

The flow, end to end

Tiered support flow A customer question is classified by the automation; tier 0 questions are answered from documents, tier 1 questions trigger a lookup in orders or deliveries, and tier 2 questions such as complaints go to a named person with the transcript and history; unanswered questions are logged for weekly review. QuestionWhatsApp, any hour Classifytier 0 / 1 / 2 · language Tier 0 · answerfrom your documents Tier 1 · lookuporders · deliveries · accounts Tier 2 · personcomplaints · refunds · "human" Replyseconds · with context Unanswered logweekly review → new answers
Three tiers, one rule: the automation never argues with a complaint.

What it needs from you

  • A sample of real conversations (an export of a week is ideal).
  • Your documents: hours, prices, policies, delivery areas, payment options. We help write what is missing.
  • Access to the systems that hold order, delivery or account status.
  • The escalation roster: who handles what, and when.
  • Your consent and data-retention wording for the first message.

Timeline shape

PhaseTypical spanWhat happens
Scoping1–2 weeksRead a sample of real conversations; sort into tiers; collect the source documents; agree hours and escalation roster.
Build2–4 weeksKnowledge base from your documents, lookup connectors, routing rules, helpdesk or inbox, reports.
Pilot2–3 weeksAutomation answers, a person reviews; unanswered questions fed back weekly.
RunongoingWeekly unanswered-question review, monthly metrics, seasonal updates (holidays, price changes).

Spans are typical shapes, not quotes. The process page explains what moves them.

What it costs

What drives the price

Tier 0 on WhatsApp from clean documents is a small build. Every lookup and every extra channel adds a piece; volume drives the monthly platform cost.

We do not publish package prices because the drivers below move the number by multiples. The pricing page lists every driver with sourced platform costs.

DriverEffect on cost
Size and state of your source documentsTen clean pages is quick; policies scattered across staff memories take longer to write down.
LookupsEach system the automation reads (orders, deliveries, accounts) is a connector.
ChannelsWhatsApp only is the base; adding email, web chat or Facebook multiplies routing.
VolumeModel usage and any helpdesk seats scale with conversations per month.
Hours and SLAsAfter-hours automation with next-morning handover is standard; 24/7 human escalation is a staffing decision, not a build.

Questions we get asked

What share of questions can be answered automatically?

We do not quote a percentage before seeing your messages. During scoping we read a sample of real conversations and sort them: answerable from documents, answerable with a lookup, needs a person. That sort is the plan and the forecast.

What happens to complaints?

They go to a person, always, with the transcript and the customer's history. The Consumer Protection Act gives Zimbabwean consumers rights we are not going to let a bot argue about. The automation's job on a complaint is to acknowledge fast and route correctly.

Can it check an order or a delivery status?

If the status lives in a system we can read (a sheet, an e-commerce platform, a courier API), yes. If it lives in a driver's head, the automation asks the driver on WhatsApp and relays the answer.

Do we need a helpdesk tool?

For one or two people a shared WhatsApp inbox is enough. Beyond that, or with email and web chat in the mix, a light helpdesk (ticket, owner, SLA) earns its place. We choose at scoping.

Will it keep improving?

Every week the system lists the questions it could not answer. Adding those answers is a short, regular task, not a project. The monthly report shows handled rate, handover rate and time to first reply.

Next step

Scope it before you book it

Fill in the scoping worksheet (five minutes, no sign-up). It produces a summary you can paste into WhatsApp or email, so the first call starts with the actual work.