Rollout playbook

Getting one person onto a new support tool takes an afternoon. Getting twenty people onto it — and keeping them there — is a change management problem, not a software problem. This is a thirty-day plan that has worked for teams making that change.

Before day one

Three decisions, made by a person, before anything is configured.

Name an owner. One person, named, with the authority to make decisions about how support runs. Not a committee. Rollouts without an owner stall at the halfway point, which is the worst possible place — two systems, a split team, and nobody accountable for finishing.

Agree what success looks like. Write it down, with numbers, before you start. "Every conversation answered within two hours" or "no conversation goes a day without a reply" or "we stop losing WhatsApp messages entirely". Vague goals produce rollouts nobody can declare finished.

Tell the team why. Not "we're moving to a new tool" but the actual reason: we are losing messages, we cannot see what is happening, customers are repeating themselves. People adopt a tool that solves a problem they recognise. They resist one that arrives as an instruction.

Week 1 — foundations

Configuration only. Nobody outside the core group is using Tola yet.

Days 1–2: channels and people.

• Connect your highest-volume channel. See channels.
• Invite the whole team as agents, set their roles, and put them in teams.
• Set channel access deliberately — decide who should see what rather than giving everyone everything.
• Set business hours honestly. What you can actually staff, not what you would like to claim.

Days 3–4: customer data.

• Import contacts and companies. See contacts and companies.
• Create the custom attributes your business actually decides things on — plan, account value, renewal date. Keep the list short.
• Spot-check twenty imported records. Phone number formatting is the usual problem.

Day 5: the vocabulary.

• Agree your labels as a team. Ten or fewer to begin with. This is a conversation, not a decision one person makes.
• Write your ten most-used canned responses. Have the agents who actually send them write them.

Week 2 — pilot

Real customers, small scope, two or three agents.

Pick one channel — live chat or your lowest-volume one — and route it to Tola for real. Everything else stays where it is.

Pick two or three agents, ideally including one sceptic. The person who is unconvinced will find the problems the enthusiasts talk themselves past.

Meet daily for fifteen minutes. What was annoying today? What could you not find? What did the customer notice? Fix things the same day where you can.

Set up basic routing. One or two workflows — route by channel, escalate your biggest customers. Do not build twenty rules before you know what the traffic looks like.

By the end of the week you should be able to say, with evidence, whether the pilot agents want to keep going. If they do not, find out why before you scale — the answer will not improve by adding more people.

Week 3 — full switch

Move the remaining channels. Do it on a Tuesday or Wednesday. Never a Friday, never before a holiday, never during your busiest month.

Keep the old system open and forwarding so nothing in flight is lost.

Have the owner and a lead watching all day. Not doing their normal job. Watching.

Announce it to customers if it changes anything they see — a new email address, a different chat widget, a changed phone number. If nothing customer-facing changes, say nothing.

Turn on assignment now that real volume is flowing. Balanced distribution, sensible capacity limits per channel.

Week 4 — tighten

The queue is working. Now make it better.

• Add an SLA so nothing sits unanswered, and so you find out where you are actually slow.
• Build the second wave of workflows based on what you have watched for a fortnight — not what you guessed in week one.
• Turn on AI copilot for agents. Drafts only, nothing sent unreviewed. See Tola AI.
• Start the help center with the ten questions you have answered most since launch. You now know exactly what they are.
• Set the shutdown date for the old system and tell everyone. Hold it.

Day 30 onwards

Month 2. Approve AI FAQ suggestions from your real conversations. Consider letting the assistant answer directly on live chat. Review your labels — some will be unused, some will be missing.

Month 3. Run your first proper report review with the whole team. Volume by channel, response times, what customers actually contact you about. Old system cancelled.

Ongoing. A monthly thirty-minute review: what is slow, what is repetitive, what is being escalated that should not be. Every answer is either a workflow, a canned response, a help center article or a product fix.

Training your agents

Agents need less training than you think and different training than you expect.

One session, forty-five minutes, hands on. Not slides. Have everyone open a real conversation and work it. Cover exactly five things:

1. Finding your queue and what the views mean.
2. Reply versus private note — the distinction that matters most and is easiest to get wrong.
3. Status: open, pending, snoozed, resolved, and when to use each.
4. Inserting a canned response and running a macro.
5. Assigning to someone else and adding a participant.

Everything else can be learned when needed. Teaching reports, custom attributes and campaigns on day one guarantees none of it is remembered.

Then: one named person to ask. Not a documentation link, a person. Adoption problems are almost always "I did not know where that was" and they are solved in thirty seconds by someone sitting nearby.

Then: leave them alone for two weeks. Resist reorganising anything while people are still learning where things are.

What to measure

Four numbers, reviewed weekly. More than that and nobody reads them.

Compare against week one, not against an industry benchmark. The only comparison that means anything is your own trend.

Why rollouts fail

Nobody owns it. The single biggest cause. A rollout is somebody's job or it is nobody's.

Everything at once. Channels, AI, campaigns, help center, custom roles, SLAs — all in week one. The team drowns and blames the tool.

Configured by someone who does not answer conversations. Routing rules designed in the abstract survive about a day of real traffic.

The old system stays open. A team with two inboxes checks both badly. Set the date, announce it, hold it.

No feedback loop. Agents hit a problem in week one, work around it, and never mention it. Six months later the workaround is policy. Ask, weekly, at first.

Declared finished too early. Getting messages into Tola is the start. The value comes from what you automate, answer and prevent over the following quarter.