AuthorAdvice.AI
Conversation design for an AI assistant that guides authors through the publishing process
Timeline
3 months
My Role
Solo project
Conversation Designer
Tools
Voiceflow
Figma
There’s so much advice available for writers looking to publish a book, but it’s hard to tell what is accurate, what is informed by real world experience, and what applies to your book in particular.
The odds for success are slim for each book, so having accurate and insightful advice can crucial.
Each book is different, and the best advice is tailored to each individual project. Using generative AI could provide that bespoke experience, but it would sacrifice the reliability and expert sourced advice of a more designed conversational experience.
Designing a conversation that focuses on accepted wisdom for general advice, while strategically using LLMs to provide more subjective feedback on things like the author’s elevator pitch.
Preview of Final Product
Let’s rewind and see all the steps I took to arrive at this result.
Conceptualizing
Why an AI assistant?
General Advice along with Tailored Guidance:
Navigating the publishing process requires a combination of broad, general advice informed by expert sources, with some guidance that’s tailored to each individual project.
An AI assistant is the ideal format for such a challenge because it can provide both.
The Freedom to Ask Many Questions:
Most people don’t have access to an industry expert, and if they do, there’s often a fear of imposing on them with too many questions.
Talking with an AI assistant about publishing takes away that concern. You get the benefit of expert guidance and you can ask as many questions as you like
Chat vs. Voice
I decided that publishing advice is better suited for a chatbot over a voice interface.
Writers will want to carefully think about how they describe their book, which doesn’t lend itself to off-the-cuff discussions.
AuthorAdvice.AI bot’s advice will naturally be complex, given the subject matter. Users will have an easier time following along if they can read through a response at their own speed, and more than once if necessary.
Hints at Delight
An early idea that was later discarded during user testing:
AI Cover Art Generator: The user could be prompted to provide a description of their work, the title, and the author’s name, and the AuthorAdvice.AI bot would generate cover art based on it. The image could be used as a motivating tool, telling the user “Picture this! You can get there!”
See below for user feedback and why this feature was not developed.
Research
Competitive Analysis
Blogs
Strengths:
Can provide advice for a specific genre.
Can draw from real experience.
Weaknesses:
Info can be outdated.
Hard to know what sources to trust.
Can contain generic advice not specific to your own project
Books
Strengths:
Info from industry professionals
Weaknesses:
A lot of info to consume
Not specific to your own project
General Purpose Generative AI
Strengths:
Tailored advice
Weaknesses:
Info can be hard to trust
Wasn’t designed with the input of industry professionals
The obligation is on the user to know what to ask, but new authors don’t know the things that they don’t know. With a designed conversation, however, the bot can guide the user to the most helpful information.
User Research
Designing
User Personas
Because the user interviews resulted in a MVP focus on the early stages of publishing, the persona created for this stage is for an aspiring writer.
Bot Persona
Product Principles:
Trustworthy and well-informed
Supportive
Intelligent Without Pretension
Note on text block length:
The user is likely in a stressful state, unsure about their chances of getting a book published, and there is so much info for them to digest.
Overwhelming them with large chunks of information can discourage people or result in them disengaging.
Instead, the goal with the voice design is to break up advice into digestible bits.
Information Architecture
Conversation Flow diagram
Sample Dialogue
Conversation Design Canvas
Sample Dialogue Transcript
Expert Rewrite
Feedback and Testing
Voiceflow Prototyping
User Testing
Here are three main examples of substantive changes:
Long Tail
Edge Cases
"Tell me the race/gender/sexual orientation of an agent."
"An author/agent relationship can be incredibly close, and ideally can last for many years. Many authors feel a desire to work with someone whose identity aligns with their own, for example. I’m unable to provide personal details about an individual agent, but you may find answers naturally through the research process.
Would you like to go over some best practices for researching literary agents?”
"Can I query multiple agents at the same agency?"
"The answer to whether you can query multiple agents at the same agency really depends on the agency’s policy on the matter. Many literary agencies will specify their policy on their website, and it’s a good idea to do your homework on the matter before sending out your queries. If you get interest from both agents at the same agency, it could lead to complications down the line if the agency had a policy against simultaneous submissions.
Would you like to go over some best practices for researching literary agents?"
"What if I don’t have a full manuscript ready?" / "When do I know I’m ready to start querying?"
"Ah yes – when is the right time to start querying. There are a few things to consider. If you’re working on a novel, you’ll need a full draft of the book ready to go. Agents rarely consider a partial manuscript, especially for a debut novel. You may hear stories of people selling their book on a partial, but you don’t want to bank on being the exception rather than the rule. The traditional advice is to finish the novel, revise it, have trusted readers provide feedback, and keep revising it until you’ve taken the draft as far as you can on your own. That’s when you start querying. The point of this advice is that you only get one chance to make a first impression, so make it the best it can be.
If, however, you’re working on nonfiction, the answer is a bit more complicated. For narrative non-fiction, like memoir, the advice is generally the same as for a novel. If you’re working on another type of nonfiction manuscript, then you typically sell that based on proposal, which usually consists of an outline, biography, platform, market analysis, and a few sample chapters. The advice about revising before submitting it is equally applicable.
Would you like to hear more about what goes into a typical query letter?"
Repair Flows
No Match
Escalating detail
Hints and discoverability
Task Failure
Combining hints and discoverability with painted door testing to validate the idea and build a list of potential users.
Next Steps
Develop: Build out more chats for later stages in the publishing process
Monetize: Affiliate links or premium features. Licensed versions with pro features for writing conference attendees or creative writing courses
Metrics: Analyze newsletter signups. Measure SUS score