comparison · 5 min read
Clicks Health, Novoflow and Amera are not selling the same healthcare agent
A healthcare procurement comparison separating provider back-office agents, computer-use EHR deployment, and payer claims infrastructure.
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
Published · Updated
“Healthcare agent” is too broad to buy against.
Clicks Health works on revenue-cycle and healthcare back-office workflows. Novoflow deploys computer-using agents inside medical practices’ EHRs. Amera normalizes incoming claims data for downstream payer and administrator systems.
All three touch healthcare operations. The procurement question is different for each.
The matrix
| Company | Primary buyer | Where the agent works | Human boundary | Public implementation signal |
|---|---|---|---|---|
| Clicks Health | RCM companies, providers, and revenue-cycle teams | EHRs, payer portals, PDFs, desktop apps, and legacy systems | Exceptions, approvals, clinical/legal accountability remain human-owned | Follows documented workflows; public site names Epic, Oracle Health/Cerner, athenahealth, eClinicalWorks, ModMed, NextGen, and MEDITECH |
| Novoflow | Medical practices, hospitals, and provider operations teams | The practice’s EHR and phone/scheduling workflow | Staff keep ownership of escalations and patient-facing judgment | One admin login; AI forward-deployed engineers map SOPs; deployment takes weeks |
| Amera | Health insurance payers, TPAs, and self-funded plan administrators | Claims intake, normalization, and structured output to downstream systems | Confirm who owns downstream decisions and controls; intake automation does not establish adjudication capability | Accepts EDI, PDFs, bills, receipts, faxes, and proprietary inputs; outputs structured claims data |
The category mistake is putting all three in a “healthcare automation” shortlist and asking which has the best model. The first question is which system is paying for the work and which system owns the data.
Clicks: start with a bounded back-office workflow
Clicks describes agents that work inside existing healthcare systems, following a team’s documented rules and routing exceptions for human review. Its current site is focused on revenue-cycle work: prior authorization, eligibility and benefits, denials follow-up, payment posting, referrals, claim cleanup, and patient-credit refunds.
That makes Clicks a fit for a provider or RCM company with a painful, repetitive workflow and a clear exception owner. The public guide says to start with work that is high-volume, repetitive, data-ready, clearly owned, and expensive. That is a better starting point than “automate the whole back office.”
There is a maturity question to resolve first. Clicks' earlier YC material described recruiting and executive-search work, while its current site leads with healthcare RCM. Ask which named healthcare workflows are live, which are pilots, and who owns configuration before treating the marketed list as deployable scope.
The public site says Clicks is HIPAA- and SOC 2 Type 2-compliant and keeps an action audit trail. Those are company-published claims. A buyer should still ask for the exact control scope, BAA terms, approval surfaces, and what happens when the agent cannot safely complete a task.
Novoflow: buy computer-use interoperability for providers
Novoflow’s wedge is different. It says its agents operate a medical practice’s EHR, deploy in weeks, and do not require staff to change how they work. The setup is one admin login followed by AI forward-deployed engineers mapping SOPs and refining the agent before go-live. Its public site also demonstrates a voice scheduling agent and says the infrastructure is HIPAA and SOC 2 Type II compliant, with a BAA available.
Choose this path when the constraint is not a single RCM task but the gap between a provider’s desired workflow and what its EHR or legacy system lets an external tool do. Computer use can avoid a long integration project. It also makes permissions, auditability, screen changes, and failure recovery more important.
The buying question is operational: who owns the login, how are changes to the EHR detected, and how does a human take over a call or workflow? “Works with any EHR” is a deployment claim to validate against your specific instance, not a reason to skip the pilot design.
Amera: buy the claims data layer
Amera is aimed upstream. Its public description says it converts messy claim inputs—EDI, PDFs, bills, receipts, and non-standard submissions—into structured data that plugs into existing systems. The company frames itself around health insurance payers and the self-funded ecosystem, not a clinic’s front desk.
That changes the procurement owner. Amera belongs in a payer, TPA, or plan-administrator conversation about intake, normalization and downstream data formats. Its homepage uses broader processing language, but the more precise About page supports that data-layer boundary. Treat validation, reconciliation, stop-loss reporting and adjudication as questions to resolve, not established capabilities from this evidence. It is not the same buying decision as an EHR agent or patient-call product.
Amera’s site reports “99%+ data accuracy across all formats” in its product material. That is a company-published claim, not an independent measurement. A serious evaluation would define the input mix, error taxonomy, adjudication rules, and the human review path before treating the number as useful.
The decision tree
- Is the buyer a provider or RCM operator with repetitive payer/admin work? Ask Clicks which workflows are live versus pilots, then scope one confirmed workflow such as prior authorization or denial appeals.
- Is the buyer a provider whose main problem is making an existing EHR do more work? Evaluate Novoflow’s computer-use deployment, login boundary, SOP mapping, and escalation path.
- Is the buyer a payer, TPA, or self-funded plan administrator processing heterogeneous claim inputs? Evaluate Amera’s intake, normalization and downstream format. Ask separately where exception handling, audit trails and claims decisions happen.
- Is the workflow clinically or legally accountable? Keep the licensed or designated human as the final owner, regardless of which agent operates the software.
For founders, the positioning lesson is the useful part: “healthcare AI agent” is not a category. The workflow surface—provider back office, EHR execution, or payer claims infrastructure—is the category. That determines distribution, security review, pricing, and the person who can say yes.
Sources — snapshots observed 2026-09-19
About the author
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.
First1000 ↗ · X ↗