Bitter but Useful Truth About Bots for Business. Part 3
Valuable guidelines on how to steel oneself for a bot launching
You won’t do it in a month
And the bigger the company you work in — the longer your project will take. Six months to develop and implement a bot in a big company are okay. Pretty often clients want to skip acceptance tests, although this is a very important stage where you can find out that new content, hypothesis or even new idea are much needed. The first business skills that we’ve built were thrown through 3–5 focus groups before becoming accessible for end-users. And there have been cases when right after these tests skills were completely rewritten.
Still, there’s no way you get some rest after the bot is out. Get ready to spend at least 2–3 months for a tuning and trial run, because a bot may stumble on something new. Get ready to spend your time monitoring it, checking on the progress status, reading the dialogs, and keeping your ear to the ground. Because voice interfaces are forever, and that’s fine.
You might need a pilot
A pilot is a must in case you have doubts as to the need for a bot, and your ability to adapt it to your business processes and topics. Especially if you don’t know what vendor to choose — pretty often platform capabilities shown in ads or tender docs differ from reality — and a pilot will show an objective picture and shaper the process of deciding on the platform for you.
For vendors, these pilot projects are a real pain. Because business doesn’t want to invest and then they sour on the technology. The business believes that the pilot is going to be free, or at least cheaper than a full-fledged implementation. At the same time they set the same KPIs they’d set for a full-scale project.
If your bot is trained on real clients and their real queries and problems — it will be able to do everything it has to during the production. The channel, the number of topics, the traffic — all that affects the cost of a pilot project.
You should have a clear understanding of the difference between a pilot project and a full-scale commercial operation. Are you planning to cover all the topics? Will you need any IT systems integrations during the pilot? Are you planning to use logs? Now or later? How long the pilot will last? What traffic it is intended for? What channels are you planning to use during the pilot? Who’s going to undertake a tech support — vendor or your employees?
A pilot insures you from the costly errors, allows you to improve on your expertise, and can be of great assistance in building an effective automation solution.
You will need some back-up from your colleagues
It is crucially important to specify who would develop and implement a bot. And I’m not talking about vendor selection, I mean your own team. There are always lots of talented people behind the bot — project managers, who are responsible for project goals and KPIs; linguists, who work on dialogs and scripts; developers, who are responsible for implementation. Altogether they are building a project to make it useful for their clients.
If this is relevant to your business, you should notify your security team right from the beginning of the bot project you are starting. It is not unusual that they mingle in the project when you are literally ready to cut the red ribbon and launch a bot.
The project must be agreed with the marketing team, especially if you are planning to use a bot in new communications channels. All the employees destined to work with a bot — should be ready (in some cases this may require business processes and KPIs revision).
You will have to learn about your target audience
There are multiple channels of communication — social media, messengers, phone and voice channel, mail — and each of them requires a different approach. Who is going to talk to your bot in the automated communication channel? VIPs or regular clients? What are their gender and age? How did they use to communicate with your brand?
You are supposed to know the audience that goes to the bot-powered channel in order to choose the right answer and communication style when developing the bot.
Obviously, a bot should be deployed across those channels where the audience is ready to interact with artificial intelligence. Quite often people ask text and voice bot “Are you human? Are you a bot?”.
It is ethical to warn your customers that they are speaking to a bot — they won’t get the impression you are trying to mislead them or get rid of them
The studies indicate that people act tolerantly when speaking to a bot, so, this can be used.
You won’t make a profit on the first day
A bot implementation has an accrual effect. KPIs may get low on the back of the queries rise even if a bot will handle a large percentage of frequent queries. Imagine, without a bot you had 10 thousand queries, and with the bot who doesn’t sleep and runs 24/7 you got 50 thousand queries. Permanent training, CRM and billing systems integrations will increase the automation level.
You’ve launched a bot and it handles 30% of all the queries? That’s great, now we can increase the percentage. Analyzing catch-all, expanding dataset, fostering integration. Upon that, it’s better to follow MVP (minimum viable product) system, take small steps while covering more topics and complicating the scenario with the new data, involving and training new agents.
There will be no good moral to the story. If you got communication channels, sooner or later you will be bent on automation. Once you are prepared you can achieve better results much more quickly.
And trust me, you will be pleased with the results!
Part 1 of this article defines the main fallacies most businessmen face when implementing a voice solution, and describes ways to get rid of them.
Part 2 tells how the value of a bot is comprised.
