When Sammalkko invested in Textio in 2023, the product pitch centered on job description debiasing: NLP analysis that flags gendered, exclusionary, or vague language before a posting goes live. That use case is real. A job description that uses words like "aggressive," "dominate," or "ninja" statistically correlates with a narrower and less diverse applicant pool. Fixing that language is worth doing.
But the reason we invested is not the debiasing feature. The reason we invested is what the product can become if the team executes correctly: the writing layer between a hiring manager's intent and the text that represents their team to the external world. Job descriptions are just the beachhead.
The Actual Writing Problem in Hiring
Hiring managers are not, in the main, good writers. That is not a criticism — it is an observation about role selection. The people who become engineering leads, product directors, and operations managers are hired for domain expertise and people judgment, not prose craft. Yet the hiring process produces substantial volumes of text at every stage: job descriptions, scorecard criteria, candidate feedback written for the record, rejection emails, offer letters with personalized framing, internal Slack updates on hiring status to the team.
Most of that text is either written carelessly (the hiring manager fires off something fast), written by HR rather than the manager (which introduces a gap between what the manager actually wants and what HR thinks they want), or not written at all (feedback left blank in the ATS because the form is tedious). The downstream costs of this writing gap are diffuse and hard to measure, which is why they persist: inconsistent candidate experience, legal risk from underdocumented interview notes, misleading job descriptions that attract wrong-fit candidates, and the slow erosion of employer brand through hundreds of small imprecision events.
Augmented writing — at its best — addresses this not by replacing the hiring manager's voice but by making it easier to write well. The distinction matters.
Why Debiasing Is a Feature, Not a Product
The debiasing use case, taken alone, has a product ceiling. Once a team has scrubbed the obvious language problems from their job description templates, the marginal value of the next debiasing pass decreases. Templates get fixed, hiring managers learn the patterns, and the tool becomes less obviously useful on a week-to-week basis. This is the product stall that companies built around pure bias-detection tend to hit.
The path beyond that ceiling requires expanding the writing scope and embedding the tool deeper into the hiring workflow. Scorecard writing is a natural extension: most hiring teams have scorecards that are either too generic to produce useful signal (five-point scale on "communication" with no rubric) or so detailed they take 20 minutes to fill in honestly. A writing assistance layer that suggests specific, role-calibrated criteria language — based on what strong scorecards for this type of role actually contain — changes the quality of evaluation data without adding time.
Candidate feedback documentation is another. In jurisdictions operating under GDPR — which includes most of Textio's European market — the documentation of why a candidate was not advanced is both a legal requirement and an operational asset for the recruiter managing the process. Getting hiring managers to write that documentation in a form that is both legally defensible and specific enough to be useful is a hard behavioral change without tooling support.
The Product Thesis We Are Watching
Elina spent four years working on enterprise product adoption problems, and the pattern she sees most in augmented writing tools is the difference between passive assistance and active workflow integration. Passive assistance — a tool that you can choose to use or not use before publishing a job description — has low adoption ceilings because the consequence of skipping it is invisible. Active workflow integration — the tool is part of the required flow before a posting goes live, or before a candidate's scorecard can be submitted — has higher adoption ceilings because the friction of bypassing it is real.
The shift from passive to active requires ATS integration. Textio has this in place with several major platforms. What we watch is whether the integrations are deep enough to make the tool feel like a native feature of the hiring workflow rather than a separate tab the hiring manager has to remember to open.
We are not saying debiased job descriptions are unimportant — the evidence that language affects applicant pool composition is solid and the legal exposure from documented bias in posting language is real, especially in Germany and Scandinavia where labor law scrutiny of hiring practices is more structured than in most US contexts. We are saying that a product whose primary value proposition is "we fix your bad language" has a different business trajectory than one whose value proposition is "we make your hiring team write well across everything."
What the Market Looks Like Right Now
The augmented writing category in HR is crowded at the surface level — there are at least a dozen tools that offer some version of job description assistance — and much less crowded at the level of genuine workflow integration and data-feedback loops. Most competitors are thin wrappers: they run a language model over text and surface suggestions. The suggestions may be reasonable, but the tool has no memory of previous postings, no awareness of which language changes correlated with stronger applicant pools, and no integration into the scorecard or feedback stages downstream.
Durable differentiation in this category likely comes from three things: quality of the training data on hiring outcomes (which requires access to enterprise hiring data at scale, not just text corpora), depth of ATS integration, and the ability to extend the product scope beyond job descriptions into the downstream writing moments of the hiring process.
The question for companies in this space is whether they are building toward those three things or whether they are optimizing the debiasing feature indefinitely. The former has a large addressable market. The latter is a useful point solution that gets acquired or displaced.
The European Context
One point that gets insufficient attention in the broader augmented writing conversation: the European regulatory context creates specific product requirements that US-centric tools often underestimate. Germany's Allgemeines Gleichbehandlungsgesetz (AGG) and the EU's Pay Transparency Directive create documentation requirements around hiring decisions that make thorough, defensible candidate feedback more than just good practice — it is increasingly a compliance requirement. Tools that help hiring managers write that documentation correctly and consistently are addressing a compliance gap, not just a quality gap. That compliance angle shortens procurement conversations in European enterprises considerably.
This is part of why Sammalkko's sector focus on European markets creates disproportionate conviction in this category. The product case that makes sense in a US market (improve your employer brand, attract better candidates) is augmented in European markets by a compliance case that is harder to deprioritize. Both are real value propositions, but the compliance angle tends to produce faster budget allocation in enterprise procurement.
The investment in Textio reflects a thesis that augmented writing will become a standard layer in the enterprise hiring stack — not because language bias is the only problem worth solving, but because the quality of text that mediates between companies and candidates has material effects on both hiring outcomes and legal exposure, and those effects compound. The winning product in this category will not be remembered for the feature that launched it.