What actually happened
At Sculpt 2026, Clay presented its next stage as a sales research and execution system. The immediately testable addition is Clay Workflows, available in beta to Clay users. Its visual canvas can branch, loop, run code, call AI agents and write results back to Clay Audiences. Teams can describe a workflow in natural language, inspect individual runs and rerun records after making changes.
Clay also showed an agentic approach to assembling sales campaigns from a stated goal. The proposed experience brings prospect research, segment creation and campaign preparation into an agent-led workflow with human review. This is an early-access capability, so its availability and behavior should not be confused with the broader Workflows beta.
The most ambitious presentation was the Inbound SDR Agent, which is still a pre-beta preview. Clay describes an agent that would qualify inbound prospects by voice, demonstrate a product in a sandbox and learn from existing recordings of strong sales calls. Interested companies can join a partner waitlist; this is not a generally available replacement for a salesperson.
Why a small team cares
Picture a five-person CRM consultancy selling automation projects to Spanish companies with 10 to 100 employees. Finding sensible prospects means identifying the right companies, checking their technology stack, deciding whether there is a clear integration need and discovering an appropriate contact. Those steps typically spread across spreadsheets, research tools and a CRM.
A Clay Workflow could research a defined batch, enrich known fields, score fit against transparent criteria and send approved records to the existing CRM. The immediate value is a repeatable qualification process. A salesperson can spend time on accounts that deserve attention instead of repeatedly collecting the same fields.
For a small team, this must be evaluated per usable prospect. Twenty-five records that produce eight well-supported targets are more valuable than hundreds of weak matches. Track enrichment cost, duplicate rate, incorrect claims, accepted prospects and eventual meetings. Automated personalization only helps when the underlying facts are right, especially when the message asserts a prospect uses a particular CRM or has a specific business problem.
Hype vs useful
Clay combines data access and orchestration in one environment, which makes the Workflows beta a legitimate test for businesses already doing structured B2B prospecting. It is still possible to spend heavily on enrichment steps that produce little sales value. Set a fixed test budget, inspect each data provider call and make failed lookups stop rather than cascade through paid steps.
Keep an approval point before any outbound campaign. AI-written messages can invent company context, and collecting or using contact data needs a documented lawful basis and appropriate GDPR handling. Neither the workflow canvas nor an AI agent removes those obligations.
The voice SDR demo is interesting because product-specific demonstrations are hard to automate. It is also precisely the part to watch, not adopt: Clay describes a future system, with pre-beta partners rather than a generally available service. The sensible decision this week is a small Workflows test with a measurable qualified-prospect outcome.