"Self-service or enterprise?" is usually asked as a company-size question, and it's really a stakes-and-stage question: how much do you need to learn before you commit, and what requirements must be true at scale? The honest answer for most teams is sequential, start self-service, graduate to enterprise terms, on a platform where that path exists.
What each model actually optimizes
Self-service optimizes learning speed: sign up today, ground the agent on public-safe content, run a real pilot this week, and let your own funnel data make the decision, no sales cycle standing between you and evidence. Its limits are the enterprise requirements it defers: SSO, SLAs, custom agents, deep integrations, procurement's paperwork. Enterprise deployment optimizes certainty at scale: contractual commitments, security review completed, governance configured, stakeholders aligned, and it costs you calendar time up front, which is fine when you know what you're buying and fatal when you don't yet. The mistake is treating these as tribes (SMB buys one way, enterprise the other) rather than as stages of confidence.
The market's failure modes, and how to read them
This category splits into vendors that offer only one path. Self-serve-only platforms demo fast and stall at your security review, no SSO, no SLAs, no certifications, and the pilot you loved becomes the tool you can't deploy. Enterprise-only platforms are quote-gated white-glove engagements: you can't try the agent on a real page without a sales cycle, which means committing before learning, on the vendor's evidence instead of yours. Both failure modes teach the same evaluation rule: the platform's range is the feature. Published pricing plus a free pilot at one end; SSO, SLAs, SOC 2 Type II, ISO 27001, and custom depth at the other; and no product switch in between.
A decision framework that survives contact
Choose the self-service start when: you're validating the category, your pilot can run on public-safe content, and your own data should make the case. Choose enterprise engagement from day one when: regulated data is in scope immediately, procurement gates any production use, or the deployment's first lane already touches core systems. Choose sequentially, which is most teams: pilot self-service in weeks one to four (the 30-day shape), run the security review in parallel, and convert to enterprise terms with evidence in hand, your funnel's numbers negotiating for you. The sequence only works if the platform spans it; check that before falling in love with either end.
The test that sorts vendors in one afternoon
Ask two questions: can I start today without talking to you, and can you pass my security review when I scale? Any vendor that answers yes to only one has chosen your buying process for you. (Disclosure: spanning both answers is deliberately how Skippr is built, free pilot and published plans at one end, enterprise requirements at the other, because the sequential path is the one we'd want as buyers.)
Questions buyers actually ask
Is self-service risky for larger companies?
Not when scoped to public-safe content while the review runs in parallel, it's the fastest honest way to generate internal evidence.
Do we lose pilot work when converting to enterprise terms?
On a single-platform range, no: the grounding, agendas, and data carry forward; the terms and governance change around them.
What if procurement insists on enterprise-first?
Then use the pilot as the evaluation artifact inside that process, a live agent on a real page beats any RFP demo script.
See what a live agent actually does
The category is easier to watch than to define. Fifteen minutes is enough to see where the mechanism differs from everything it gets confused with.