How and Why to Add a CuratorBot to Your Content Strategy
For content curators, adding a CuratorBot on Facebook Messenger is a no-brainer. Get one in front of your followers before everyone else.
(This is a significantly edited repost of Messing about with bots (alpha), published on my Curation Hub late last year. All related posts here.)
It’s not often that a new media comes along — I’ve only seen it a couple of times in my lifetime — so over the Christmas break I created my first chatbot. I did it to get a handle on the form, and I’ve been alpha-testing it on friends ever since. I only faced one small problem…
I’m not a developer. I’ve been working with developers since I started building database-driven websites in 1995, but I can’t code.
This is exactly why chatbots interest me so much. It took 5 years of wide web availability before tools like Blogger and Wordpress brought online publishing to people with no technical skills. This transformed the web and paved the way for social media. However, when chatbots appeared over the horizon, they were followed almost immediately by tools like Chatfuel, which promise to ease the pain for code-challenged botmakers like myself.
If chatbots succeed, the resulting changes could be equally as profound, expecially for my field of public communication and online communities. Here, I suspect bots could deliver public services more effectively than websites or apps.
Thus, I wanted to find out about bots and blog about my findings for Chatbots Magazine.
What’s a CuratorBot?
The clue is in the name: it’s for people or organizations who bring value by helping others find useful stuff.
Brings value by helping others find useful stuff.
This, of course, applies to any organization with content they’d like people to access where they are (cue obligatory graphic for all bot posts, left).
In my case, I built my bot with Chatfuel on top of my Curation Hub, which I launched a few years ago on Tumblr. To understand “HubBot,” you need to understand my Hub.
My Hub is the content store from which my bot dispenses content relevant to the user’s request
My Hub is the content store from which my bot dispenses content relevant to the user’s request. Over 90% of the 2650-and-counting posts on my Hub is “Stuff I Like” — i.e., annotated summaries of and links to stuff that I’ve read and recommend to others. These are published automatically as part of my daily productivity routine, where I annotate, tag and store resources to my personal library on Diigo (briefly: any resource I tag ‘like’ on Diigo gets autopublished as a Tumblr post on the Hub via IFTTT).
And the rest? I post to 3 blogging platforms, so I bring all my blog posts together under Stuff I Think, as well as summarise my work (Stuff I Do).
As each resource is also tagged by topic, a Tag Menu such as “bot” (left) will link to resources I’ve tagged Like around the Web; my own posts (including this one); and work examples - all relevant to bots.
My Hub is thus central to my personal content strategy. Around 18 months ago I added my “Top3ics” enewsletter to that strategy to ensure I was really absorbing the stuff I was reading and annotating every morning, before writing a (hopefully) original post of my own (more: Stop Drowning in your Inbox).
My CuratorBot extends this curation service to a new medium.
My CuratorBot simply extends this service to a new medium.
To do that, it fetches three resources I’ve Hubbed relevant to whatever keyword I give it, plus a link where I can find more (i.e., the Tag Menu for my keyword).
CuratorBots Are More Than Newsletters
It also incorporates a bot-version of my newsletter — the three very best stories from the 20 or so curated in my latest Top3ics enewsletter are presented in a dedicated block, which I can broadcast to all subscribers via Messenger as often as I want.
To begin with I called it Top3ics too, but tests showed that confused people, so it’s now called High3lights. It has to be updated manually (for now at least), but that takes literally three minutes.
Although most enewsletter managers would kill for the Open Rates bots often get, however, creating a NewsletterBot is nothing special — it’s just transferring a content product from one medium to another.
a NewsletterBot is just transferring a content product from one medium to another
What’s unique with CuratorBots is that instead of subscribing to my newsletter — and getting 20 or so resources I choose to send, when I choose to send them - users can simply ask, at any time, for whatever interests them.
All with one thumb, as they wait for the bus, or a job interview, or a sales call, or a date. Assuming we have similar interests, they’ll find something to read, given that I feed my Hub practically every day.
users can simply ask a bot at any time for whatever interests them, all with one thumb as they wait for the bus
So it’s a small but significant addition to my personal content strategy (left).
And now for the good news: HubBot was ridiculously easy. Here’s how.
Build your own CuratorBot
You will need:
- an online database of curated content (i.e., each piece of content is tagged) which provides an RSS feed for each tag
- a Facebook Page (Messenger bots can’t be linked to personal Facebook profiles)
- a Chatfuel account.
And a large piece of paper to sketch out the botflow
I was pretty unimpressed with the (free) tools out there for sketching out botflow — i.e., creating the bot-equivalent of a sitemap — so after an initial sketch on paper I ended drawing it up with Visio. Here’s my current flow:
If you think this is overkill — that you can probably make do with scribbles on paper — don’t.
You’ll tweak and tweak and tweak again
A comparison with the December original shows how this has changed in a few weeks. You’ll tweak and tweak and tweak again as your first users get lost, ask questions you didn’t anticipate and make useful suggestions. If you don’t maintain an editable botflow — digital, whiteboard, whatever — you’ll end up losing them, and possibly yourself.
Walkthrough: User Input and RSS
Technically this bot is simple — the most complex thing it does is take a keyword from the user and plug it into an RSS module to pull up three relevant articles from my content store.
For example: let’s walk through Welcome, the first experience the user has (see screenshot, above). It has 3 blocks:
- textblock 1: expectation management (important — see below)
- textblock 2: two buttons, with the “…” in the text leading the user to an invitation in the …
- … User input block, which invites them to feed the bot a keyword, which is stored in the User Variable ‘Topic’.
- This variable is then fed to the RSS import block:
These screenshots, btw, are from my contributions to the Chatfuel community. At that point I was pulling resources from my Hub on Tumblr, but that meant the user was being sent to my Tumblr, when they presumably would prefer to go straight to the full article. It also meant that they weren’t getting any pictures, as my Diigo-IFTTT-Tumblr process doesn’t handle them. Since then I’ve changed the RSS input block to pull the content from my personal Diigo library.
(2020 update: I just replaced my Tumblr with a MyHub.ai account, solving the problem — my bot grabs the curated content from MyHub, but sends the user to the full article)
Conclusions from alpha testing
CuratorBots are easy, Faceted search would be awesome
CuratorBots like HubBot are incredibly easy if you already have a content strategy underpinned by a content database with an RSS feed per tag.
But what is easy is not always what is best. What CuratorBots really need is faceted search. But that does require coding, and so is for my next post. Follow me to make sure you get it.
Expectation management is vital
Alpha testing showed the importance of expectation management.
Opening a conversation with “give me a keyword and I’ll get you some reading” was just stupid of me — I saw users asking for “Eagles”, “Massive” and “Japan”, none of which I could help them with. It’s a big language.
“A curatorbot’s welcome message should tell users what topics it curates”, says Captain Obvious
So if you’re building a CuratorBot, make sure your welcome message gives users some idea of what sort of topics it curates!
Don’t assume familiarity, nor that anyone reads
This, of course, will resonate with anyone who’s ever designed a website and then watched UX tests. Basically:
- don’t expect everyone to be familiar with bots
- don’t expect them to read your text blocks properly
Unfortunately, these are often in opposition to each other.
One alpha tester, for example, didn’t realise he could click the buttons, so he retyped the button content.
Solution: AI rules for each button and block to catch (and hide) as many of these user errors as possible (left).
But I also decided to include ‘Click the button’ instructions in each text block. Unfortunately, this lengthened the texts, even though I could see that alpha testers were skipping over these text blocks, missing important info.
I still don’t know what the ideal text block length is, nor which form of grammar (neutral text, commands! or questions?) is best for buttons.
But building HubBot, of course, is exactly how I’ll learn.
This was my first of several instalments of the Adventures of a non-Coder in the World of Chatbots. If you got this far, Recommend it & Follow me for more. Try out the bot at https://m.me/freshintegral.
Related reading
- You can contact me and subscribe to my enewsletter, as well as check out everything I Like, Think and Do, at MyHub.ai/@mathewlowry.
- On MyHub.ai: more stuff about bots in general, and mybot in particular
- Align your personal productivity routine with your personal content strategy and Stop Drowning in your Inbox
- Update: here’s my followup post: Boosting CuratorBots with Faceted Search
