Bot Project  -  to Be or not to Be?

 

IDEA and REASON

My thought process was pretty simple, “..why do it on Google Forms when I could create a ChatBot!” – that was back on Monday 19th November 2018; as I sit here typing this post, I think I’ve gone too far down the chatbot-shaped rabbit hole. In short, I think I’ve over complicated it!

The motivation behind the chatbot, was to create an interactive survey as part of my on-going research into the job interview and my attempt to answer one question – “Has the job interview format changed that much?”

Why am I interested in interviews?

The reason for this is twofold; one, there has been a proliferation of technology that has been created to support, automate and replace the ‘human element’ of the interview process. Secondly, the traditional interview focused on skills and experience are also now starting to incorporate (or be replaced) by the assessment of soft skills, behavioural and motivational drivers and, the often referenced, ‘future potential’.

In addition, I wanted the survey to be a ‘community project’ as I intend to share the results with everyone that requests a copy of the report; in short, if you’re a Hiring Manager, In-House Talent Acquisition Professional, Agency Recruiter, HR Professional or you work for an RPO, the report could inform your plans for 2019. Also, as an interviewee, this could provide you with valuable insight into the formats you might experience during the interview process next year.

 

The BOT

Like most non-techies, I used Chatfuel to create my Bot and I started with the usual ‘Welcome message’ and choices of options for the end user basically giving them the option to start the survey, be persuaded by me to take the survey or completely opt out (see slide show below).

This slideshow requires JavaScript.

Then I had used the Google Forms template to create the sections of the conversation tree and options for the person interacting with the Bot – the difference between the Forms and a Bot is that you have to put a lot more thought into the flow because you’re no longer just asking someone to ‘tick boxes’. Now you’ve reached the start of the process of constructing your ‘Groups’ and/or ‘Sequences’ which is both a wonderful option, however, I had to consider the ‘experience’ of the person now.

I think this was the point that I started to layer numerous options for two reasons; I was thinking about what possible answers and options a person might want to see, also, I was thinking about the options I’d like to see and experience.

AdobeStock_169179236 copy.jpeg

The AIM

The survey basically aims to answer the following questions:

  • What has been your role within interviews during the course of 2018?
  • What is your industry sector or your clients industry sector?
  • What interview formats have you used in 2018?
  • What interview questions have you used in 2018?
  • What interview technology have you used in 2018?
  • What interview formats, questions and technology are you planning to use / use more of / implement in 2019?

: however, with the Bot at my fingertips, I started to expand the details of the question(s) to see if I could add more value to the survey results (the thereby, add more value for everyone who wanted a copy of the report afterwards).

The QUESTION

This is the point I am at now….do I continue with the Bot or do I stick to the Google Forms option?

Sources:

GIF courtesy of Sylvia Boomer Yang @boomeryang

 

 

 

 

 

 

 

Advertisements

JeezusBot — The Pivot (Part 3)

JeezusBot Image 2

Three days of action and the results

Tuesday 20th June 2017

Before my three days of action started, I had a lunchtime meeting with a Product Manager and friend; he’s currently transforming a London-based vehicle hire firm by delivering their enterprise app offering. This chap knows his Product design (*and he’s got great recommendations for new music); so the focus of the meeting — Who? What? Why?

These principles may be common place for you Product professional folks, however, to me this resonated as my process of simplifying JeezusBot was about answering a simple question of his ‘best use’.

Wednesday 21st — Friday 23rd June 2017

The simple answer to the ‘Who?’ question is; everyonethe ‘Why?’ I can answer by saying — I believe it’s possible to eliminate the use of the common CV.

Eliminating the use of the common CV was the reason I created JeezusBot; he was to act as a virtual curator to help individuals learn more about me and contact me. Essentially, I was exploring the value of a ‘virtual CV’ that was more interactive and personal in terms of engagement. I now think that this would have to be more of a platform-play / marketplace idea and it would need scale to execute properly. Additionally, the interaction with JeezusBot was quite low, hence the need to pivot.

Action 1: Search for data and insight

A constant through this pivot stage is the requirement to gather more data and insight; for this reason, I decided to run a promoted Twitter campaign to understand where the chatbot conversations are happening and who is having them. This campaign ran throughout the 3 days of action (21st-23rd June) and I actually promoted an original blog post I’d written about the creation of JeezusBot and the results where as follows;

  • Tweet Impressions = 11,147
  • Tweet Engagements = 163
  • Engagement Rate = 1.46%
  • Clicks = 69
  • Media Engagements = 93
  • Top 4 countries for Impressions = Portugal (6,500), France (2,798), Ireland (1,306) and Germany (291) — the UK came in 5th place (143) followed by the USA (109).
  • Top 4 Keywords Impressions = AI (9,720), Persona (4,833), Artificial Intelligence (477) and then finally chatbots / chatbot (204)
  • Top 4 Interests Impressions = Comedy; Movies & Television (7,485), Music (7,389), Action & Adventure (6,289) and Comedy; Hobbies & Interests (6,279)
  • Impressions Female/Male = Male (5,910) and Female (4,521)

Conclusions

  • Portugal is a country where chatbots are a subject to conversation
  • The UK is not a country that appears to be part of the chatbot conversation at present
  • Artificial intelligence is clearly a hot (super-hot!) topic
  • The female/male impressions regarding the topic were that far apart — a positive sign!
  • Comedy and Hobbies & Interests — not entirely sure what to make of this insight at this stage; however, perhaps chatbots are perceived as still ‘social’ as opposed to ‘business orientated’?
  • My content (blog post) wasn’t that interesting or my expectation is too high given the resulting engagement rate

Just as an FYI to anyone wondering about the inclusion of the keyword ‘Persona’ into the campaign, I added this because of the company Persona who build chatbots for celebrities such as Katy Perry.

Action 2: Exploring Alexa Skills

As part of my research into the new user journey when looking for a new career opportunity, I’ve registered to develop a ‘Skill’ for the Amazon Echo voice recognition assistant, Alexa.

I’m now in the process of starting talks with a developer to guide me through the more technical implications of creating a ‘Skill’ for the platform; at this stage, there appears to be no other options for job search advice or options.

Action 3: Networking & connecting

Interestingly, a meeting this week has resulted in the suggestion that JeezusBot becomes a ‘social’ character (as opposed to functionally for business); and it’s linked to one of my personal interests; graffiti art. Having checked Botlist just before posting this blog, there hasn’t been a chatbot created specifically for graffiti art.

I’m also in the process of setting-up a call with the CEO of a recognised technology solution provider to discuss a recruitment chatbot proposition; this is a business related conversation, however, it does present an opportunity to learn more about the landscape of chatbots for talent acquisition.

Summary of the 3 days of action

At this stage, it feels like the 3 days past and I didn’t achieve a huge amount; however, that is subjective as my expectation was perhaps misaligned. If I look at it objectively, I have more insight that will shape my thinking about target markets in terms of geography, areas of interest and the current hot topic of the moment — AI. In addition, I have another avenue to explore in more detail for the use of JeezusBot — graffiti art.

So what next? — I need to set myself a ‘re-launch’ date!

JeezusBot — The Pivot (Part 2)

Image-1

Community feedback from the first 48 hours

After the decision to pivot my chatbot, JeezusBot, I took two direct actions with the first being;

  1. Conduct quick poll on Twitter and share it on Facebook and LinkedIn to gauge sentiment regarding the social and practical usage of chatbots.

As you can see (below) the 4 options loosely grouped into ‘social’ and ‘practical’ actions; the ‘Provide interview advice’ option was obviously in-line with my agenda;

 

The result? — a whopping (??) 9 responses from the Twitter poll (*thanks to those of you that kindly responded); and the over-ridding sentiment was that a chatbot would be useful if it could ‘Resolve a payment query’ (see below)

 

Drawing conclusions from this 48-hour poll is quite simple:

  • Polls via Twitter don’t seem to deliver much engagement or my community didn’t find the question compelling enough; I’d prefer to know which one is true
  • Sharing the poll via Facebook and LinkedIn had zero effect; although, on LinkedIn it had 415 views apparently (??)

My second direct action was to…

2. Review and dissect the data collated from all interactions and promoted posts for JeezusBot

This was slightly more insightful in relation to the promoted content that I’d paid for on Facebook as the ‘Insights’ section on FB is pretty solid and will give you actual data to review in relation to each post;

The post this data refers to is one entitled “Alexa and me sitting in a tree…” that had the following blog post attached to it ‘The future of your search for a new career’. The blog post is based on outlining my thoughts about how we will search for new career opportunities in the (very near) future; a cross platform approach that combines voice-recognition (Amazon Echo/Google Home), chatbots and mobile.

From the data (above) the blog posts target age group (18–24 year olds) are firmly in the ‘millennial’ camp with a 3:1 ratio of men to women, and interestingly, from a geographical perspective Portugal and Israel came above the UK and Ireland and other European locations. Are these my ‘early adopters’? — more research required and this is a valid action point.

Next steps

I need a more sizeable group of users to get data on the preferred use of a chatbot — Twitter and Facebook promoted posts could be a solid option; and research into the use of chatbots by the millennial demographic utilising communities and forums such as Botlist could be good starting points.

The next phase of this sprint cycle will run until Friday 23rd June 2017.

JeezusBot — The Pivot (Part 1)

Press to Reset

When your Bot is not delivering any value, it’s time to pivot!

The realisation

JeezusBot was introduced to the virtual world back on the 15th February 2017 and you can read more about his early story here and here. Since the 24th May 2017, the date of the last interaction with him, he has been a state of idling for two reasons; nobody is interested in him and I didn’t know how I was going to refresh his content to make him useful and/or interesting.

I needed some inspiration…or direction.

As is often the case when I personally think about ‘inspiration’ I think about something complex, large scale and transformative — basically, I over complicate everything! Thankfully, as I was reading through my daily content regarding all things chatbot, I happened upon a series of interviews conducted by Carina Skladal entitled BotXperts and a light came on — thanks so much Carina!

The inspiration and direction I was looking for was — simplicity. I needed JeezusBot to actually offer one thing that would be of value to people that interact with him. This started a brainstorming session that resulted in a number of options, however, a question kept nagging at me — ‘what is valuable?’. My perspective is one (and only one) view of what could be valuable; I needed more input from a community of people to define the use cases.

The first actions

The next steps were quite simple;

  1. Conduct quick poll on Twitter and share it on Facebook and LinkedIn to gauge sentiment regarding the social and practical usage of chatbots
  2. Review and dissect the data collated from all interactions and promoted posts for JeezusBot

The next steps

Wait 48-hours to review the results of the poll from Twitter and take further actions. Simple!

The future of your search for a new career?

Robot Wars

Starting your job search through a voice recognition assistant

Talking to a ‘robot’

Amazon’s Echo and Google’s Home are currently nestling somewhere between the ‘Innovators’ and ‘Early Adopters’ on the Product Adoption Curve illustrated below; if you have one of these two products you will know the excitement and anticipation that comes with setting-up and then discovering what your ‘assistant’ can do…you may also know what they cannot do (which is also evident at this stage). On balance (once your expectations have been realigned), the voice recognition assistant is a useful addition to the home (in my humble opinion).

Product Adoption Curve

For those of you that view these devices differently, you likely will be part of the ‘Early Majority’ or even ‘Late Majority’; essentially, you want to purchase a tried and tested version 3.0 or 4.0 of the product that has been optimised without any of the frustrations. As was the way with individuals that came late to Twitter, Facebook and even Snapchat, the benefits of these platforms will have been fully realised and utilised by others (*in all these cases, ‘others’ means a growing Global community).

In this context, would you want to be at the back of a Global queue for your next career opportunity?

If the answer is ‘no’ (and I suspect it might be), I would encourage you to start getting ‘socially comfortable’ with speaking to a voice assistant; as with Siri, we still seem to be self-conscious about asking a ‘robot’ to do something for us in front of our family, friends and colleagues (*unless they are tech geeks like us).

Who Made You Robot?

Let the people speak

I am in the process of conducting a survey at present simply entitled ‘How do you look for a new career opportunity?’; early results (as of this morning) still illustrate a traditional approach to the job search. When asked how individuals typically look for a new job and which platforms and tools they use, LinkedIn still comes top with referrals from friends and colleagues following closely behind. Other results show the amount of time dedicated to looking for a new career opportunity on a daily basis; surveyed individuals allocate 30 minutes (53%), followed by 1-hour (20%) and then 2-hours (6%) and then ‘Other’ (21%) to the daily search.

Other interesting data so far illustrates an interest in applying for career opportunities via Facebook and Snapchat whilst WhatsApp is also a channel individuals would like to utilise as part of the application process.

The survey says…

While the survey sample is still small and the survey will be open to more participants until the end of May 2017, the early data indicates the following:

  • Individuals may have a limited amount of time per day to search and apply for new career opportunities
  • LinkedIn is a primary platform
  • Facebook is one of the social platforms that individuals would consider using to search and apply for new career opportunities
  • Individuals would be open to using messaging channels such as WhatsApp and potentially Facebook Messenger
  • Individuals have expressed a potential interest in having a mentor to support them through the job search process
  • A significant percentage would use a chatbot to assist them with applying for opportunities

The future model?

Bringing us back to the opening subject of voice recognition assistants and their integration into the way we will search for new career opportunities; look at the (rather basic) process flow below and tell me if this is something you would or wouldn’t consider; I welcome all constructive and respectful thoughts and opinions.

Gifly Image

7 early user learnings from JeezusBot

Chatfuel Persistent Menu

And what happens when you disable the ‘allow user input’ option?

In answer to the above question, f**k all in the case of JeezusBot. I turned off the ‘allow user input’ button at 10.01am on the 24th April 2017 and since then, nothing has happened — and you know why?- because JeezusBot either offers zero value or…he offers zero value.

Retracing my steps in terms of creating JeezusBot there has been some interesting learnings;

  1. User engagement — unless you know your target audience, expect limited interaction; even friends, family or work colleagues will not necessarily offer you a steady flow of interested users as their understanding or levels of interest are not either provoked or met at the surface view level .
  2. Facebook promoted posts — make sure your posts are targeted and not just the default settings as the target demographic is to vast. You don’t have to spend a lot to learn this, I spent £5.00.
  3. Gamification and humour are important to retain user engagement and drive more interactions through a singular conversation. Also, remember that your own personal humour may not translate to a wider and diverse audience.
  4. The option to contact the ‘creator’ of the bot has limited value; users appear to want to interact with the bot only.
  5. Users are interested in the intelligence level of the bot and will often have a preconceived idea about the richness of the engagement; if that expectation isn’t met, it can lead to immediate user dissatisfaction.
  6. Remember, you could have your bot ‘Blocked’; invariably, some users may have one interaction with your bot and then decided they are no longer interested. Be prepared for rejection — new tech doesn’t always equate to interest if poorly executed.
  7. Repetition is not a great user experience; if your bot cannot answer a question and/or query and simply defaults to the ‘Welcome message’; chances are your user will either try another conversation branch or give up — the latter is more likely than the former.

The journey continues….

JeezusBot Main Pic

JeezusBot — Birthing of a chatbot

JeezusBot Intro

JeezusBot was introduced to his virtual world on the 15th February 2017 to a limited fanfare; he simply introduced himself. On the 4th March 2017 he had his first interaction with an individual based in Greece; she seemed intrigued, however, after checking if he could speak Greek, asking him ‘why don’t boys love me’ and asking him to ‘talk to her’, she gave up as JB clearly didn’t have the right answer (*or any answer) to any of those questions and couldn’t interact in a more complex conversation.

Lessons learnt by both JB and me from this interaction — he was far too one-dimensional.

Build

JeezusBot was built on the Chatfuel platform as I am a non-techie and it’s easy to construct the Bot structure and conversation blocks (see below); the premise of JB was to introduce people to me — this may sound like narcissism, however, that wasn’t the intention — my thoughts were that it could be used to test the idea of a virtual CV as it would act as an introduction to me as a person, thereby creating a more 3-dimensional view of me as an individual for potential new employers.

Chatfuel Image

With this idea in mind I created specific blocks for ‘Welcome message’ which is a standard part of the structure and then created blocks for Career, University, Blog Posts, Interests (*outside of work) and a Contact me option. Inserting a mixture of images and GIF’s to provide visual content was important as I thought text alone would have people tuning out immediately — much in the same way as CV’s might do now.

Measure

After inviting a small group of friends and respected colleagues to JB’s page via Facebook from my own FB page, while also posting to a popular chatbot group called Bots; I asked individuals to interact with JeezusBot and let me know what they think. That was on the 13th March 2017 and by the 14th March 2017 JB had been through 5 interactions;

  1. 10 conversational responses and a GIF in response
  2. 8 conversational responses and a compliment in response
  3. 4 conversational responses and a question — ‘Are you smart?’
  4. 2 conversational responses
  5. 12 conversational responses and a question about religion

In addition to posting personally, I paid for a promoted post on Facebook on the 16th March 2017 which was a YouTube video of a street performer playing drums; the idea was to provide JeezusBot with more of an identity through having an interest. The post reached a total of 336 people (308 paid for and 28 organically); this resulted in only 7 post engagements and zero traffic to JeezusBot’s Facebook page.

Facebook Image

Learn

Through these interactions I drew some observations which are as follows;

  • Your target audience(s) need to be defined; create personas for each one you wish to engage with and align the content accordingly
  • Learn from each interaction and implement an iterative design process
  • Humour increases engagement and interaction (*as long as the Bot knows when to respond with the correct response and not a random “A Zebra” — see above pic)
  • If an individual is familiar with you or the brand, they will have a very different expectation in terms of the information they can access; in this case, they will not be satisfied with high-level or introductory information as they’ll seek deeper more meaningful information and data
  • Assumptions will be made about your Bot; for example, that it might be able to speak multiple languages, that it can understand contextual conversation, that it has a level of intelligence beyond its actual capacity
  • The Bot’s novelty factor may only last as long as one interaction if there is no reason or value in returning for another interaction / conversation
  • Not everyone has an interest or can quantify the value of Bot technology at this stage as it perhaps hasn’t penetrated their work or social world
  • Allocate a budget and design a plan of activities for promoted marketing material linked to your Bot; one-off promotions don’t drive consistent traffic to our Bot’s site
  • Repetition of information reduces the quality of the user experience when an individual interacts with your Bot; users seem to forgive the limited capacity of your Bot on one or maybe two occasions
  • Immediate choice is more appealing to the user; a question and decision to make engages the user more quickly than a extended introduction

JeezusBot Main Pic

What next for JeezusBot?

At this stage, I’m planning to reboot JeezusBot taking into account the aforementioned points I have learned and create an extended platform of information that he can offer and access for the individuals that wish to interact with him.

Any constructive and positive suggestions and/or feedback are always welcome.