A practical evaluation guide

Before you add another customer-service tool, ask these questions.

Hiring more people, adding a chatbot, buying contact-centre software or introducing another communication tool can all be reasonable choices.

The important questions are what each option solves, what it leaves behind, and who or what will coordinate the work between your knowledge, channels, systems and people.

Preview your Chatterbox

Go straight to the five tests

Before you compare options

What needs to work better for your customers and your team?

Different problems require different solutions. Before comparing products, establish where customer needs are being lost and where administration is being created.

  • What happens when a customer calls after hours?
  • What happens when a call drops or a customer leaves a queue?
  • What happens when a customer replies to an outbound message?
  • Can customers communicate in the language they are most comfortable using?
  • Can your team see what the customer wanted, what has already happened and what must happen next?
  • Which work is repeated because conversations, knowledge and systems do not connect?
  • Which requests can be resolved safely, and which require human judgement?

If the problem is unclear, adding another tool may simply move the administration somewhere else.

Most organisations try one of these

Each option solves a legitimate part of the problem.

The question is not whether an option is useful. The question is whether it can carry the whole customer need through, or whether your team must continue coordinating the gaps around it.

1Keep going as you are

What it preserves. Familiar systems, familiar processes and no immediate implementation effort.

Questions to ask

  • Where are customer requests currently being lost?
  • Who follows up on missed calls, abandoned queues and after-hours contact?
  • Who handles replies to outbound messages?
  • Can you see why customers contacted you and what needs to happen next?
  • How much time is spent repeating, forwarding, capturing and checking work manually?
  • Is the current approach working, or is the organisation simply accustomed to it?

What may remain unresolved. The missed requests, fragmented context and repeated administration remain.

2Hire more people

What it adds. Human understanding, empathy and judgement.

Questions to ask

  • How many additional people are needed to cover peaks, queues and after-hours demand?
  • Which South African languages can the team support consistently?
  • How will people find the right approved answer while the customer is waiting?
  • How will answers remain consistent across people, shifts and locations?
  • How will each conversation be captured in a form the organisation can use?
  • Who will review what happened and determine what needs to happen next?
  • Does hiring remove administration, or add more people who must perform it?

What may remain unresolved. Cost increases, availability remains finite, and customer knowledge may still sit in calls, inboxes and individual notes rather than becoming useful organisational memory.

3Add a chatbot

What it adds. A fast, familiar way to answer straightforward questions on a website.

Questions to ask

  • Does it respond only from your approved knowledge and rules?
  • Does it know when it does not know?
  • Can it understand what the customer means, rather than match a keyword or menu option?
  • Can it work beyond the website through voice and SMS?
  • Can it remember what happened across later interactions and channels?
  • Can it take an appropriate action, or does it stop after producing an answer?
  • Can it hand over to a person with the conversation, relevant knowledge and next step attached?
  • Can it work within South African language, accent and vernacular realities?

What may remain unresolved. It may answer within one interface while people still coordinate the customer need across other channels, systems and teams.

4Buy contact-centre software

What it adds. Powerful channel, queue, workforce and contact-centre management when the organisation needs that operating model.

Questions to ask

  • Do you need a new contact-centre operating model, or help with a smaller set of customer-service problems?
  • What infrastructure, licensing, integration, training and organisational change will be required?
  • Does the platform understand the customer need, or primarily route the interaction?
  • Can it use your approved knowledge consistently across channels?
  • Can it understand and respond within South African language realities?
  • Does it remember the need across calls, messages, systems and later contact?
  • What happens after the interaction reaches a queue or a person?
  • Can you prove the specific use case before committing to the wider transformation?

What may remain unresolved. A contact-centre platform can manage an extensive operation, but may be more transformation than the problem requires. The intelligence needed to understand and carry the customer need through may still need to be added.

5Add a dialler or messaging tool

What it adds. An efficient way to reach many people through calls or messages.

Questions to ask

  • What happens when the customer answers or replies?
  • Can the tool understand the question, explanation or objection?
  • Can it respond using the correct approved knowledge?
  • Does it know when a person is required?
  • Can it remember earlier contact and avoid starting again?
  • Can it determine and record what needs to happen next?
  • Who handles exceptions, follow-up and replies that do not fit the original campaign?
  • Does the tool reduce administration, or generate more responses for people to process?

What may remain unresolved. The tool reaches the customer, but does not necessarily understand, remember or carry the resulting customer need through.

6Use menu trees and voicemail

What it adds. Cheap deflection, published options and a recording of the call.

Questions to ask

  • How many callers abandon the menu before they reach a person?
  • Does the recording capture what the person needs, or only that they called?
  • Who listens to the recording, and how soon?
  • Can urgency, language and a next step be taken from the contact without a person reconstructing it?
  • Is a voice message in a browser the same as voicemail, or does voicemail only apply when a real telephone number is involved?

What may remain unresolved. Callers abandon trees under load. A recording carries no structured need, importance or next step, and it routes to nobody until someone has time to listen.

7Build it yourself

What it adds. Full control and a system shaped exactly to your operation.

Questions to ask

  • How long before the first customer is served on approved knowledge?
  • Who will own language coverage, conversation memory, handover and ongoing operations?
  • What governance is required before a customer-facing answer is allowed?
  • Can you prove the specific use case privately before the development programme starts?
  • What happens when the team that built it moves on?

What may remain unresolved. Control is real, but language coverage, governance and operations all arrive before the first customer is served.

The missing layer

Who or what will coordinate the whole customer need?

Channels carry calls and messages. Systems perform actions. People provide judgement. Knowledge and policies set the boundaries.

At the centre

  • Your Chatterboxthe coordinating layer

Connected to it

  • Approved knowledge
  • Customer channels
  • Operational systems
  • Memory of what happened
  • Allowed actions
  • Your people
Your Chatterbox is the coordinating layer at the centre, connected to your knowledge, channels, systems and people.

Something still has to:

  • understand what the customer is asking;
  • find and use the correct approved knowledge;
  • recognise what it does not know;
  • decide whether to answer, act or hand over;
  • remember what happened;
  • make clear what needs to happen next; and
  • keep the customer need moving across channels, systems and people.

If people perform this coordination, the administration remains. If fixed workflows perform it, they struggle when customers speak naturally, change the subject or present something unexpected. If an intelligent interface only produces answers but cannot remember, act or hand over, it remains a chatbot.

Your Chatterbox connects the conversation to the next step, using your knowledge and keeping your team informed.

Before it speaks for you

Establish what it can do, what it knows and where it should stop.

Before it interacts with customers, a customer-service intelligence should pass five practical tests.

1 Can it communicate through real customer channels?

Can it work through the channels your customers use? Where the job needs telephone or SMS, can it make and receive calls and messages, understand them and respond?

It should not merely transmit a call or message. It should be able to participate in the conversation on your organisation's behalf, within the boundaries you set.

2 Can it understand the need and know its limits?

Can it understand what the customer is asking, formulate a response from your approved knowledge and rules, and recognise when it does not know enough to respond?

When uncertainty, risk or human judgement is involved, can it stop and hand over to a person with the relevant context attached?

3 Can it work within South African language realities?

Can it understand and translate written conversations across the 11 official spoken languages?

Can it understand and speak naturally in South African English, isiZulu and Afrikaans, using an appropriate accent and vernacular?

Language support should be tested with your customers' real questions, terminology and ways of speaking, not accepted as a generic feature claim.

4 Can it prove itself before you commit?

Can you test it privately, using your own knowledge, at no cost and without commissioning an IT project, adding infrastructure, building integrations or requiring developers?

Can you inspect its answers, test its boundaries and rule it out quickly if it is not a fit, before it touches customers or changes your operation?

An evaluation should need minimal organisational involvement and no budget allocation or procurement to test. If it is not a fit, you rule it out early with nothing to unwind. If it has value, you take working proof into the leadership conversation instead of slides.

Testing the intelligence should be simple. Live phone numbers, system integrations and operational workflows can be introduced later, only when the use case requires them and the organisation is ready.

5 Can your team operate and govern it?

Can it be administered through familiar tools such as email and a secure portal?

Can your team monitor activity in real time, control knowledge and rules, review handovers, and use dashboards, analytics and insights to understand what customers need and what happens next?

When the capabilities work together

It becomes more than another customer channel.

Working alongside your team, your Chatterbox can:

  • handle missed calls;

  • engage customers who would otherwise remain in a queue or on hold;

  • respond to inbound messages;

  • make and handle payment-reminder calls and messages;

  • continue conversations when customers reply;

  • remain available after hours;

  • communicate across supported South African languages;

  • capture what happened and what needs to happen next; and

  • hand over what needs a person, with the context attached.

It understands the need, uses the right knowledge, knows when a person is required, remembers what happened and keeps the work moving.

Your team sees what the customer needs, what has been answered and what should happen next.

One intelligence across the work

It does not replace every tool. It connects the work between them.

A phone system carries a call. A messaging tool carries a message. A contact-centre platform manages an operation. A CRM stores records. People apply judgement.

Your Chatterbox sits between your approved knowledge, customer channels, operational systems and people. It determines what can be answered, what can be acted on, what must be remembered and what needs human judgement.

You can connect separate tools for calls, messages and records. The remaining work is understanding each request, keeping its history and deciding what happens next. Your Chatterbox carries that work through and hands over when a person is needed.

Questions to ask

How will you judge the value of the job?

  • What triggers a customer contact?
  • How often does it happen, including peaks and after-hours demand?
  • What does answering and following up cost your team today?
  • What does a missed or late contact cost you, using the measure that matters to your organisation?
  • Who will act on the questions, issues and replies your Chatterbox captures?
  • What will you measure to decide whether the job is worth the fee?
We would rather be ruled out

It should prove that it fits before you invest in it.

Worth testing if

  • customer needs disappear after hours, in missed calls, abandoned queues or unhandled replies;
  • outbound calls or messages create responses your current tools cannot handle;
  • customers struggle to communicate in the language they prefer;
  • your team repeatedly searches for answers, captures the same information or forwards work without context;
  • your knowledge exists, but is difficult for customers and frontline teams to use;
  • you need an intelligence layer that works with existing systems rather than replacing them; or
  • you want to prove a use case privately before committing to operational change.

Probably not the right solution if

  • you only need to broadcast one-way calls or messages and do not need to handle responses;
  • your existing systems already understand, remember and carry customer needs across channels, systems and people;
  • the problem is simply a temporary shortage of people and human handling is the intended long-term operating model;
  • you do not have, or are not prepared to identify, approved knowledge and rules from which it may respond;
  • you are not prepared to define when it should stop and hand over to a person;
  • you need a replacement for your CRM, billing system, PABX or complete contact-centre platform; or
  • there is no specific customer-service problem against which its value can be tested.

If none of these conditions applies, your Chatterbox may not solve a problem worth buying it for.

Prove the fit

Do not take our word for it.

Give your Chatterbox a webpage or approved document and test it privately. Ask real questions. Change the language. Test what it knows. Test what it does not know. See whether it answers correctly, stays within its boundaries and hands over when it should.

  • No customer exposure
  • No new infrastructure
  • No integration project
  • No developers
  • No cost for the private Preview
Preview your Chatterbox Speak to us about a use case

If it cannot prove its value using your own knowledge, rule it out.