As a startup we have a lot of energy, a lot of ideas and a lot of fun, but what we don’t have is a lot of time and enough resources. Now that our platform is getting momentum, we need to scale up our team, but as efficiently as possible. So we decided to use our own product to solve this.
The best way to design a bot like this is to analyse the current process and especially zoom in on the conversations you would have with candidates to decide whether to hire a person or not.
Analysis of the current hiring process
So one thing we would normally do when hiring a new team member is to (1) post an ad on our social channels and await response. Once we get a response (2) we would try to find the social profiles (3) of this candidate, on Linkedin or Facebook for example, and use this for a first selection. This selecting (4) is crucial if we don’t want to waste too much time doing interviews that lead to nothing. Again, our time (and probably yours too) is very precious.
Next we would plan an intake interview (5) with the candidates that seem interesting based on their profile. To not waste too much time of our team, this intake would typically be done by me, not waste any valuable time of my team members yet. During this intake (6) we would typically ask the candidate to elaborate on certain aspects of their profile and qualify their fitness for the job based on questions that would depend on the specific job.
After this intake we want to assess the results (7) and compare with the results of the other candidates to make a short list. We will invite (8) all short listed candidates for a second interview. This interview (9) will be held with the rest of the team too, since they obviously need to work with their new team mate after hiring. After the second interviews the team will decide (10) so I can continue hiring or apologising (11).
Design of the desired hiring process
So how would we design a bot that can do this better?
First of all, we would place the bot, i.e. the URL to the bot, in our Ad. This way, candidates can self-service their way into our hiring process. This saves step 2, the collecting responses step.
Secondly, we decided to start the bot by asking the candidates to login using either their facebook or their linked-in accounts and asking them permission to read it. This is very easy to do using Botsquad since both platforms support the OAuth authentication standards, which are supported out of the box, see development section of this article. So this saves step 3 of the normal process.
Then we would just skip steps 4, 5 and 6 of the process by letting the bot continue the process by asking a custom set of questions that are selected based on the job specifics and candidate profile. This saves both us as the candidates a tremendous amount of time. The candidate doesn’t need to take the time off and travel to our office.
The bot is then able to ask questions that verify the experiences expressed in the social profile and/or ask questions that would help the team assess the fitness of the candidate for the job.
Now we have all data we need to do a qualitative assessment of the candidates by the team based on much more data would normally have in step 4 without us having to spend a single minute up until this point. The data will be collected in a very structured and disciplined way (by the bot) which makes it very easy to compare candidates: so called apples-with-apples-comparison.
So we decided to leave step 7 a manual, human step, since team dynamics are a key succesfactor of our company (and I guess any company for that matter). The team will decide, not a bot nor a computer system in general will do this. Also, the data will be presented as it is collected. It is still up to the candidate to present him or herself the best way he or she can, as well as it is up to the team to analyse the data, have a human conversation with the candidate and then make a decision based on the data, emotions and gut feeling: love to be a human :-)
This would leave step 7 to 11 as our primary steps in the hiring process which cannot be removed.
So this means we need to cover steps 2 to 6 in the chatbot and make sure we ask the right questions so the team can perform step 7 in a qualitative way. Then, if you think you are there, you need to compress it as much as possible taking into account the valuable time of our potential candidates. No candidate will spend an hour talking to a bot to get our attention. It needs to be as quick as you would expect from a typical chat conversation.
We identified the following almost generic questions to be most valuable during interviews:
- Why do you think you are a perfect candidate?
- Do you have at least 2 former colleagues or clients that will confirm this?
- What work are you most proud on so far?
- What advice would you give us based on what you’ve seen and know so far?
- Did you already use our platform before?
- If so, what is your impression so far
For developers we added:
- Do you have experience with chatbots?
- If so, which technology did you use?
These questions gave us most valuable information on the quality of the candidates. Not only the specific answers, but also how it is answered is relevant to us.
Because we are looking for people that will actually add to the team we bet quite heavy on questions 4 and 6. We want new ideas and realistic critics on what we are doing and how it can be improved.
For Linkedin we are still waiting on permission for the V2 API’s so we can actually read the (former) positions of the candidate so we can ask more precise questions about that. In the mean time we ask people to elaborate on their tag line, which we can read using the V1 API’s.
On top of the above we ask which languages the candidate speaks if we cannot get it out of their profile, their birth date and to describe briefly their personality. And since there are actually people that made an effort out of writing a good resumé we included an upload CV function :-)
Using this bot, we save ourselves and our candidates hours of valuable time that can be spend better!
Next article I will explain to you how we implemented this bot using Botsquad.