AI Was Supposed to Make Recruitment Cheap. Here’s Why It Might Make It Worse

AI recruiters can now do the whole thing. Find any job from across the open web. Apply for you. Source candidates from across the open web. Screen them. All at ultra low cost. So recruitment should get dramatically cheaper, right?

Probably not. Response rates on open channels are collapsing under the weight of mass personalised outreach, and candidate data is getting harder to trust. AI recruiter startups are focused on building closed-loop platforms that own both trusted candidate data and a direct, reachable relationship with users. Closed-loop platforms need strong network effects, so the AI recruiter market is likely to consolidate around a handful of winners (or even one) with huge pricing power. So recruitment costs may never come down.

The only way to stop AI recruitment collapsing into a few giants is interoperability built on a neutral, on-chain candidate data layer. Velocity Network Foundation is the biggest attempt to build that layer so far, but its permissioned design and centralised governance undermine claims of neutrality. A shared layer only works if no entity can control it or change the rules. That’s the kind of layer Kablio aims to help build one day.

Job search used to be constrained by fragmentation and manual work: roles were scattered across platforms, and jobseekers had to search site by site (and within each site) and write time-consuming applications. Now AI recruiters can search the whole internet for roles, surface the best matches, and apply (or make applying effortless) on almost any platform.

On the employer side, candidate data had been fragmented and hard to search or filter. However AI recruiters can now pull profiles from across the web, personalise outreach and run screening interviews at scale.

In other words: AI has turned search, outreach, and screening into internet-wide, low-cost actions. That sounds great, but when everyone can message anyone, open channels like email and LinkedIn get swamped. When anyone can effortlessly claim anything about themselves or invent a persona, misinformation explodes. The open web starts to crumble.

AI recruiter startups are focused on building closed-loop platforms that can reduce spam and maintain more trustworthy data. The problem for users is that AI recruiter platforms like Kablio and Jack & Jill are network-effects businesses: they’re most valuable when everyone’s on the same platform. That dynamic will drive consolidation into a handful of winners (or even one) with huge pricing power.

What is a blockchain? It’s a public database that’s (often) tamper-resistant and equally open to all. That neutrality is what makes it powerful: anyone can use it knowing no single entity can rewrite records, restrict access, or hike fees.

Sure, a private company could build a shared database and undercut a blockchain on price. But nobody would trust it, because once it achieved serious network effects, the company could raise access fees or change the rules in its own favour. And that’s why blockchains are so great: the rules will only ever be set to serve the network, and everyone knows it.

So how would on-chain credentials work? Say I worked at American Express. Amex could publish a signed statement on the blockchain confirming my employment, linked to my address and signed by an official Amex address. Or the claim could come from the address of a reputable verification agency that has confirmed my employment with Amex.

The first benefit of putting a credential on-chain is public verifiability. It isn’t self-reported. It’s an immutable public record issued by an employer and independently verifiable by anyone (or only by parties I choose to grant access to).

The second benefit is portability. If I switch AI recruiters, I do not have to re-prove my history from scratch. The new AI recruiter can verify the same credentials instantly, which makes switching easier.

When credentials are publicly verifiable and portable, users do not need to all live inside one closed AI recruiter platform. They can be spread across multiple interoperable AI recruiters and switch between them easily, which keeps the market fragmented and competitive.

The biggest attempt so far to build an on-chain career data layer is by the Velocity Foundation. It has built a protocol it describes as public, open, and trusted, giving individuals self-sovereign control of their credentials while allowing employers and education providers to issue and rely on immutable, verifiable records.

The problem with Velocity is that, despite having a shared protocol and participant governance, the network is permissioned by a non-decentralised governance body. That makes it non-neutral. And without credible neutrality, it doesn’t make sense for their data to be on-chain.

Issuers on Velocity (organisations certified to issue verifiable credentials about a holder) are approved by the Foundation’s Registrar, and node operators (it runs its own L1) are restricted to “trusted institutions” drawn from the Foundation’s members and issuers.

So the question is: why go on-chain if you’re going to permission the chain anyway? Blockchains are slow, complex databases with cryptographic overhead. Public blockchains use public-key cryptography because open networks need a way for strangers to prove control and authorise writes without a central identity provider. But if every participant is already KYC’d and approved, you can just use normal accounts and permissions (e.g., email + standard auth).

Their only plausible defence is that permissionless is still the endgame. But their own docs say they abandoned this goal after regulators raised concerns.

So unless they’re secretly planning to become permissionless, and/or hoping for a regulatory change that makes it viable, there’s little reason for Velocity to be on-chain.

The Foundation may have good intentions, but it is governed by a consortium (its founding members include big corporate heavy hitters like ZipRecruiter, Oracle, SAP, Randstad). And consortium-owned networks, even when built for the industry, rarely stay neutral once the value from network effects kick in. Visa is the classic case: it started as shared bank infrastructure, then became a tightly controlled, profit-driven network.

That’s why I’d be wary of building Kablio’s data layer on top of Velocity. We’d be at the mercy of a consortium, and there’s a significant risk the network gets privatised and starts hiking access fees.

I’d build a Solana-based protocol designed to become immutable and truly decentralised over time, where anyone can issue credentials. Trust would come from the guarantee that neither I nor any other entity can change the rules for our own benefit.

So how would you trust issuers of credentials? Issuers announce their public key, and an open ecosystem can emerge to assess reliability, much like the web’s broader reputation layer. Anyone can issue credentials, but trust is handled in the open: verifiers decide which issuers to rely on, instead of deferring to a central registrar.

Credential data lives off-chain in a decentralised store as encrypted records controlled by the candidate. They can share records directly, or use zero-knowledge proofs to confirm specific facts (for example, “I hold this licence”).

The protocol would rely on Solana validators for consensus and security, with no central party deciding who can participate at the infrastructure layer, as in Velocity.

The protocol is funded in USDC and a native token: employers and recruiters pay to verify credentials, and each payment is automatically split between network fees and issuer rewards, keeping issuance free and charging only for verification.

Of course, this is just a high-level sketch. There’s a lot more to work through, including governance, security, account recovery, privacy, dispute resolution, incentives, and plenty more besides.

A shared, credibly neutral credential layer is the cleanest way to stop industry consolidation. It lets candidates carry verified history between competing AI recruiters, and it lets employers and recruiters verify claims without buying into any one platform’s walled garden. Velocity is aiming for that outcome, but its permissioned design risks creating a new gatekeeper.

Even if the industry believes Velocity will stay neutral, I don’t see how it kickstarts the network. A shared data layer is more likely to emerge the other way round: a successful AI recruiter builds an on-chain credential layer for its own product, commits to it fully, then opens it up and makes it permissionless once it’s useful. That might look like benevolence, but it’s also rational self-interest. The pioneer benefits from a larger network, higher engagement, and upside in the protocol’s token economics.

This is an idea I dream about a lot at Kablio. But the protocol design is genuinely hard, and bootstrapping it is harder. We’re early-stage, and we can only do one thing well at a time. The plan is to earn the right to come back to this: get real traction with the AI recruiter, learn what credentialing needs to look like in real-world workflows, and then, if we’re in a position to shift industry behaviour, take a serious swing at building the neutral layer the market actually needs.