Public principles
How Talkfolio Operates
Talkfolio is a credibility interface for professional hiring. It exists because static CVs leave recruiters to infer scope, ownership, and fit from dense text — and because AI tools that optimise CVs for keywords or polish reduce the signal recruiters need.
Talkfolio takes the opposite stance. The principles below are public so that anyone evaluating Talkfolio — recruiter, candidate, journalist, investor — can verify what we will do and what we refuse.
The shift
A recruiter does not read first.
They ask.
The system is designed to reduce time-to-confidence — the gap between opening a profile and knowing whether to make contact.
What we are
A profile that can be questioned.
Experience is structured with claim types, scope boundaries, and linked evidence. An AI assistant answers questions using only data the candidate has published.
The AI is an interface, not the source of truth.
Gaps remain visible.
How it works
Claim types
Every claim is one of: owned, delivered, designed, influenced, proposed, observed. Scope is part of the claim, not optional. These are not labels. They determine how claims are interpreted and how the AI answers about them.
Evidence or flag
If a claim has linked evidence — recommendations, case studies, public artefacts — the AI can cite it. If it does not, the claim is published, but visibly flagged as stated experience without external corroboration. Evidence is never invented and its absence is never hidden.
Bounded answers
Recruiter-facing AI answers pass through a deterministic filter before the language model is called. If the profile does not support an answer, the AI says so. It does not infer beyond evidence. It does not generate in the candidate's voice. It does not soften gaps into prose.
Visible gaps
When evidence is partial or absent, the answer says so explicitly. The system tells the recruiter where evidence is strong, where it is partial, and where it is absent. Not a score. A map.
Candidate owns publication
Nothing is visible without explicit publish. Editing a published claim returns it to draft.
What we refuse
No scores. No match percentage. No fit rating. No candidate ranking. These compress evaluation into a single signal that obscures real judgment. We do not compute them.
No AI-generated assertions in the candidate's voice. The AI answers questions about a profile. It does not write claims, generate testimonials, polish phrasing, or insert experience the candidate did not approve.
No persuasive UI. No gamification. No completeness scores. No behavioural nudges.
All input methods are equal. File, free text, voice, image, and LinkedIn export each produce the same structured output. Voice is not a beta feature or an advanced option.
No stale published content. Editing a published claim returns it to draft. The recruiter sees what the candidate currently stands by — not what was true six months ago.
What we measure
The primary metric is contact rate — the proportion of profile visits that lead to recruiter outreach. Not profile views. Not time on page. Not message volume.
Candidate engagement is not measured in any way that would push candidates to spend more time in the product than the product warrants.
Two things we will never do
We will not introduce candidate ranking, scoring, or matching against vacancies. The recruiter, not Talkfolio, decides who fits.
We will not retain candidate data beyond what the candidate has explicitly published or saved. Voice recordings, drafts, and abandoned uploads are deletable, ephemeral by default.
Violations should be visible. The feedback widget and productcare@talkfolio.io are in the footer of every page.