Where AI fits, and where it doesn’t

We believe creative work should stay human.

We also think technology can be genuinely useful around that creative work, particularly when it takes some of the administrative load off literary agents, helps them find the stories they’re looking for faster, and helps writers get responses sooner.

That’s the role AI plays in QuerySort.

Our position

AI is a complicated subject in publishing, and we understand why.

Like you, we don’t want AI writing books. We don’t want it replacing literary judgement, or deciding which stories matter.

But we also don’t think every use of AI is the same.

Software can use AI for one specific function without handing the whole workflow over to it. A product can decide what the AI is allowed to process and what it never sees.

We’ve been very deliberate about drawing those lines in QuerySort.

Brooke Hahn and Kayley Greenland, Co-founders of QuerySort, who support human made creative work

So what does AI actually do?

QuerySort uses assistive AI for one specific, bounded task: comparing an agent’s wishlist with the information writers provide in their queries. It then organizes those queries by how much they align with the wishlist and provides a short explanation of why that alignment was identified.

At its core, it’s comparing language with language.

What has the agent said they are looking for? What has the writer said their manuscript is about? Where do those things align?

It doesn’t decide whether the writing is good, whether the story deserves representation, or whether an agent should request pages. Those decisions belong solely to agents.

What QuerySort can do is make a busy query inbox easier to navigate, so agents can spend less time sorting and more time actually making those decisions.

QuerySort compares the agent's wishlist with the writer's query. The agent is seeking quiet, character-driven literary fiction set in small coastal or rural communities, centred on women in mid-life. The query is for a literary novel about a fifty-four-year-old lighthouse keeper returning to a Cornish fishing village. The wishlist comes directly from the agent, and QuerySort never accesses the manuscript or sample pages.

What QuerySort doesn’t do

We’ve been very deliberate about where AI stops in QuerySort. These boundaries reflect what we believe technology should, and shouldn’t, be doing in publishing.

QuerySort does not:

Read manuscripts or sample pages, including pages attached to an email or pasted below the query itself (e.g. picture book manuscripts written in the email body)
Assess writing quality
Decide whether a query is “good”
Automatically reject writers
Make representation decisions
Replace an agent’s judgement
Generate or alter writers’ creative work

Why that matters

The problem we’re trying to solve is an administrative one.

Agents can receive an enormous number of queries and have very limited time to work through them. On the other side of that inbox are writers who may wait months to hear back.

If QuerySort can make that inbox easier to navigate, agents can find the queries that look most relevant to what they’ve said they want sooner, and writers can hopefully get responses sooner as well.

We believe technology can be used for that.

The problem we’re trying to solve is an administrative one.

Agents can receive an enormous number of queries and have very limited time to work through them. On the other side of that inbox are writers who may wait months to hear back.

We believe technology can be used for that.

Before QuerySort: one inbox with 1,284 unread queries, and no way to know which match your wishlist until you read them. With QuerySort: the same inbox, sorted into 220 high-alignment, 346 medium and 718 low-alignment queries, all still there and readable. Agents know which queries to read first, and writers can hear back sooner.

Why use AI at all?

Because wishlists aren’t lists of keywords.

Agents describe what they’re looking for in all sorts of ways. They might be eagerly seeking something, curious to see more of it, open to it without actively looking for it, or very clear that it isn’t for them. Writers, meanwhile, may describe a closely aligned story using completely different language.

All of that context matters when QuerySort organizes an inbox.

Simple keyword matching can tell you whether the same words appear in two places. It can’t reliably understand what an agent means by them.

Language models are useful here because they can interpret the meaning and intent in natural language. That allows QuerySort to understand an agent’s wishlist as a whole and compare it with the information writers provide in their queries.

That’s why we use AI here.

Because wishlists aren’t lists of keywords.

All of that context matters when QuerySort organizes an inbox.

Simple keyword matching can tell you whether the same words appear in two places. It can’t reliably understand what an agent means by them.

That’s why we use AI here.

What happens to query data?

QuerySort temporarily processes information from an incoming query so it can compare it with the wishlist the agent has provided to QuerySort.

Once that processing is complete, the query content is deleted from our limited processors.

QuerySort does not store or retain a database of writers’ queries.

Manuscripts and sample pages are not processed at all, whether they’re attached to an email or included beneath the query itself.

QuerySort temporarily processes information from an incoming query so it can compare it with the wishlist the agent has provided to QuerySort.

Once that processing is complete, the query content is deleted from our limited processors.

QuerySort does not store or retain a database of writers’ queries.

Manuscripts and sample pages are not processed at all, whether they’re attached to an email or included beneath the query itself.

What about model training?

This is one of the concerns we hear most often around AI in publishing, and it’s one we take seriously.

Writers should be able to submit unpublished work to an agent without wondering whether that material is quietly being used to train an AI system.

QuerySort does not use writers’ query or manuscript content to train AI models.

This is one of the concerns we hear most often around AI in publishing, and it’s one we take seriously.

Writers should be able to submit unpublished work to an agent without wondering whether that material is quietly being used to train an AI system.

QuerySort does not use writers’ query or manuscript content to train AI models.

We really care about this

These boundaries aren’t something we added because AI is a sensitive topic. They’re part of the way we wanted to build QuerySort from the beginning.

We’re two women who saw a very real problem in publishing and thought we could make it better. One of us, Brooke, is also a querying children’s writer and is deeply immersed in the publishing world (and may possibly fan girl internally whenever she speaks to an actual literary agent).

So the concerns around AI and creative work aren’t abstract to us. We care about them deeply and personally.

At the same time, we’ve both spent much of our careers working in software and see every day how useful AI can be when it’s applied to a narrow, practical job.

That’s really our position: human creativity matters enormously to us, and we think technology is at its best when it supports the work around that creativity, rather than trying to replace it.

Read Brooke's article on her AI perspective below:

I don’t want AI writing stories either. But assistive AI can help literary agents in this very narrow, administrative way. →

These boundaries aren’t something we added because AI is a sensitive topic. They’re part of the way we wanted to build QuerySort from the beginning.

So the concerns around AI and creative work aren’t abstract to us. We care about them deeply and personally.

At the same time, we’ve both spent much of our careers working in software and see every day how useful AI can be when it’s applied to a narrow, practical job.

That’s really our position: human creativity matters enormously to us, and we think technology is at its best when it supports the work around that creativity, rather than trying to replace it.

Read Brooke’s article on her AI perspective below:

I don’t want AI writing stories either. But assistive AI can help literary agents in this very narrow, administrative way. →

Ready to sort your inbox?

Join agents on the early access list and be among the first to try querysort.

Join the waitlist