At STRV, we’ve recently implemented three chatbots – and now, our practical experience is yours for the taking. If you’d like an intro to chatbots or how they can help your company, check out Part 1 and Part 2 of our chatbot series.
Cast study 1 – FAQ bot
Candidates are always going to have tons of questions about your company, but they won’t always pick up the phone or talk to a recruiter. Your FAQ bot is here to provide answers 24/7. Test our FAQ bot.
We’ve partnered with jobpal, a recruitment chatbot company that has a ready-made solution that can be deployed almost instantly. However, this doesn’t mean we could escape doing our homework – the chatbot still needs to be fed with loads of info.
At first, we had to create the initial batch of questions and answers. Jobpal helped us by providing the most frequently asked questions, like – how big is the company? Can I work from home? Do you have set working hours? Can I bring my dog to the office?
After two months of usage, our FAQ library contains 53 topics and 460 questions.
For every given topic we had to create and record at least 5 different questions. For example, to check whether we’re a pet-friendly company, questions can be asked in the following ways:
Q: Can I bring my dog to the office?
Q: Are you dog-friendly?
Q: What is your pet policy?
Q: Can I bring a cat?
The second step is to draft the answers in a way that they would match the different questions. This means for example to avoid “yes/no” answers, because “yes, you can bring a dog” won’t logically answer someone else’s question of “what is your pet policy?”
So the correct answer is “we are a pet-friendly company.” Or, because we value our tone of voice, our answer is: “Animals are totally fine with us and our offices. Give them the full tour and impress them with our space!”
Training the bot
Jobpal has an administration panel where you manage the FAQ library. You can see different categories (topics) and edit the answers or even add in possible questions.
If the bot becomes confident in answering questions, it won’t bother you. If it’s not sure, the question goes to the training section where the admin responds with the right answer or creates a new topic if needed.
Luc Dudler, CEO of jobpal explained that at the beginning, the bot usually needs help with every second question. But the more the bot is used, its accuracy increases and can answer up to 75% of all questions.
Our monthly statistics keeps us updated about the bot’s usage and lets us see every single conversation in the messenger window.
The clear advantage of the bot is that people can contact him anytime and can re-engage in a previous conversation. It’s a pretty great addition to our current communication channels when it comes to the company and the recruitment team.
Our bot also has the ability for job searching and collecting job applications. But it seems that only a few candidates actually use it for job applications. This is an area we’ll need to work on.
After investing so much time into training the bot, we want our candidates to know about it and use it. Currently, we’re using the Messenger plugin to open a chat window on our career site. We’re also introducing our chatbot via links in candidate email templates, in our newsletters and email signatures.
We’ve also been experimenting with different Facebook advertising formats. With one option, Facebook ads won’t take candidates to the career site. Instead, they directly open the chatbot conversation in messenger.
You can’t expect a human level of understanding yet. The bot needs to follow certain patterns and it’s possible to confuse him. Of course some people do it intentionally to test its limits.
The FAQ bot isn’t a plug-and-play tool. You need to invest time in its training – by creating the list of topics with several possible variations of questions, drafting the right answer, checking with your copywriter, and validating questions that the bot can’t answer.
We aren’t able to yet say how much time our chabot saved or how many new candidates we have thanks to it. But we can definitely confirm that it creates interest, engagement and a superior candidate experience.
Case study 2 – GDPR bot
Would it be possible to collect the GDPR consent in a cool and engaging way? That was the task for our GDPR chatbot. (Test the bot)
Every month, we send a newsletter with blog posts, event invites and job updates to over 8,000 contacts in our candidate database. With GDPR taking effect, we had to collect the consent from some of them.
In May 2018, users were being bombarded with GDPR requests. We wanted to stand out and do things differently, so we’ve partnered with Feedyou in the development of a “GDPR bot”, that collects consents in a cool and interesting way.
Bots can be funny
Technically, the bot is quite simple. It’s a linear conversation: “What’s your email? Do you want info about A? Do you want info about B? Here’s the summary, please approve.” It took 3 weeks from the initial briefing to the release.
What makes the candidate experience unique is the funny and friendly tone of voice combined with cool gifs that Feedyou prepared. The experience has already started with an initial email that was totally different to what other companies were sending when requiring the GDPR consent.
This is really important. The chatbot is just a tool. To get the maximum out of it, you need emotions: curiosity, entertainment, and fun.
When you select a chatbot provider, technical expertise isn’t the only priority. Check how the vendor understands the business, marketing and copywriting. Ask for references.
We’ve confirmed that people were 3x more likely to use the bot than filling in a form.
From this point of view, ROI worked. But branding was an even bigger thing and that’s why we’ve decided to do it.
Many candidates sent us positive feedback that our GDPR process was totally different to the other emails they’ve received. Also, we’ve gained some additional visibility by sharing the case on LinkedIn – and that’s worth the investment.
Example 3 – Knowledge chatbot
Have you ever tried to find information quickly in company policies? Usually, it’s a damn nightmare. But what if you have a chatbot that does it for you? You simply ask what you want to know and it’ll sweep through company guides and immediately give you the right answer.
Talla is an intelligent knowledge base that you feed with all your guides and policies. What makes it unique is that you don’t have to access the knowledge with a standard search box, or searching in chapters. Instead, it has a chatbot feature. You can ask her questions and Talla finds the right answers. IT can be linked with Slack and therefore employees can easily communicate from there.
“I need a new phone.”
“I’m going on a business trip.”
“What are the benefit offers.”
Talla usually gives several options, and after confirming the requested one, it shows the extract of the policy. The user can eventually open the whole document and also confirm whether the information was helpful or not. This helps to train the bot even further.
Similarly to previous cases, good preparation was required. We had to collect all policies and annotate them for Talla so that the algorithm can better understand the structure of the text. Also, we had to think about possible synonyms and keywords that people could ask.
This was a huge, one-month cleaning exercise that allowed us to update our policies, remove duplicates and fill in knowledge gaps. In the admin system, every policy has an owner, and we can even set review periods for when the owner is requested to check and approve the policy.
Getting everyone’s buy-in before the start was also very important. Luckily, the benefits are so obvious that it wasn’t difficult at all.
Talla is so simple to use and it brings real value to the employees. They can just pick up their phone, type the question and get the answer. Goodbye ugly intranet, shared folders, obsolete policies and employee frustration. No doubts Talla can have a great role in onboarding new employees.
KEY LEARNING POINTS
If you plan to go ahead with chatbots in your company, consider the following points:
- Select vendors with relevant experience in making Recruitment/HR chatbots. Technical expertise only isn’t enough.
- Be ready to invest your time. The higher the complexity of your bot is, the more inputs from the employer are required.
- Good copywriting is super important. Involve a professional.
- Focus a lot on marketing. Use chatbots to increase your fame and candidate experience.
- Don’t expect miracles. Afterall, chatbots are still a relatively new tech that needs further evolution. On the other hand, as an early adopter, you’ll get a step higher compared to the rest of the market.
2 reakce na „Chatbots in Recruitment, Part 3: Practical tips“
Vies co by ma zaujimalo?
Ako by to bolo s cestinou… Predstavujem si, ze tie varianty otazok a odpovedi musia byt brutalne a velmi obtiazne jazykovo zvladnutelne.
Ci sa mylim?
Diky za odpoved.
není to tak těžké. GDPR bot (case 2) je lineární. Tam není třeba žádný language understanding a může být klidně v češtině. Mrkni na stránky Feedyou.agency – mají plno případovek v češtině.
U té jedničky – FAQ bot – to je trochu těžší, ale už také existují knihovny pro češtinu, takže to není nemožné.
Knowledge bot je jenom v AJ, ale brzy něco podobného vznikne i pro další jazyky.
Obecně – technicky to je v pohodě řešitelné, jde jen o to, aby se někomu do toho chtělo vzhledem k velikosti (malosti) trhu.