Sourcing: Dead or Alive? Reviewing SOSU 2026

Executive summary

SOSU EU 2026’s central message was that sourcing is not disappearing, but its traditional, manual form is increasingly automatable. AI can take on much of the work of finding, researching, and contacting candidates. The durable value for talent teams is shifting toward business and talent strategy, relationship-building, data and market insight, AI-enabled operations, and judgment in difficult or ambiguous searches.

The panel also cautioned that automation can amplify problems: candidate fraud, incomplete or biased data, over-reliance on platform filters, and mass outreach that turns into spam. Teams will need to redesign workflows and responsibilities—not simply add another tool.

Key business takeaways

• Automate repeatable search work; redirect human effort. Several speakers argued that database searching, routine profile screening, and initial outreach are likely to be automated. Sourcers who only perform those tasks face the greatest risk.
• Move talent teams closer to business decisions. Chad Horn said ServiceNow is shifting sourcers away from filling individual requisitions and toward building talent pipelines ahead of demand. Their market insight can inform hiring profiles, compensation, benefits, and placement strategy.
• Treat relationship-building as a differentiator. As AI makes outreach easier to scale, candidate trust, genuine networks, and domain expertise may become more valuable. Marcel also recommended searching an organization’s existing ATS and reconnecting with known candidates before defaulting to external platforms.
• Expect fewer roles, but more technical and strategic work. Guillaume argued that recruiting demand is less elastic than engineering demand: automating hiring does not necessarily create more hiring needs. Speakers anticipated smaller talent teams, with higher expectations for business judgment and technical capability.
• AI results depend on data quality and access. Panelists warned that platform filters, incomplete profiles, and visibility rules can hide strong candidates. One speaker said a LinkedIn Recruiter search returning 20,000 results may show only 1,000; the remaining profiles and platform activity patterns can distort the available pool.
• Candidate verification is becoming an early-funnel concern. Chloe described rising cases involving fabricated identities, identity theft, and rented personal data. Detecting fraud requires shared responsibility across recruiting, hiring managers, interviewers, and security—not just reliance on a tool’s risk signals.
• Avoid a race to automated spam. If every recruiter can contact thousands of candidates, outreach volume may overwhelm candidates and damage response rates. The panel’s countermeasure was to compete on relevance, relationships, and credibility rather than message copy alone.

Frameworks and practical playbooks

Redesign the sourcing workflow around human–AI collaboration

1. Give the agent a clear role brief and requirements.
2. Use AI for research, candidate identification, and personalized initial engagement.
3. Keep a human in the approval loop for candidate fit and next steps.
4. Track the funnel and review where the agent’s recommendations succeed or fail.

This was illustrated by MetaView’s Filmore product, which operates in Slack and presents candidates for recruiter review before scheduling.

Shift from requisition response to proactive talent strategy

• Build pipelines and talent communities before vacancies arise.
• Share market intelligence with business leaders.
• Use that insight to challenge or refine hiring criteria, compensation, and role profiles.
• Maintain relationships with hard-to-hire talent instead of relying only on job postings.

Build a candidate-fraud process across the hiring team

• Train recruiters, sourcers, hiring managers, and interviewers to recognize different red flags.
• Combine signals rather than treating one indicator as proof.
• Partner with security teams on escalation and verification.
• Account for this as additional work in the recruitment process.

Develop as a builder, not just a tool user

• Test tools against real workflows rather than adopting them because a vendor claims they work.
• Learn enough to configure, integrate, or create solutions that address business problems.
• Connect technical capability with recruiting knowledge and business context.

Make ongoing learning routine

Alla Pavlova recommended reserving time for self-education; she sets aside 30 minutes per week. Her closing advice was to learn a tool regularly—suggesting a tool a month as one possible pace—and use free learning resources or AI assistance to get started.

Product example and reported metrics

MetaView representative Elise described Filmore, an autonomous AI recruiting agent that works in Slack. A recruiter provides the brief; Filmore researches potential candidates, generates personalized outreach, and returns candidates for human approval. The company said a Teams version was on its roadmap but gave no release date.

Reported figures from MetaView:

• Filmore completed 8 hires in one quarter and was credited with one-third of the company’s sales hires in that period.
• The company estimated Filmore completed 86,000 hours of recruiting work in the prior 30 days, using roughly 21–25 agents for research.
• One software engineering hire reportedly took 13 days from search start to accepted offer; the candidate replied to the first email in 27 minutes.
• A senior demand-generation marketing hire reportedly took 30 days to close.
• Alla said her own candidate response rate was approaching 90%, which she attributed to the distinctive, relationship-focused element of her outreach.

These are speaker- and company-reported examples, not independently validated benchmarks.

Advice for talent professionals

• Build business curiosity, AI fluency, and analytical skills alongside recruiting expertise.
• Learn to translate business problems into workable data, tools, and processes.
• Develop relationships and domain knowledge that cannot be replicated by simply searching a database.
• Test AI tools on actual workflows, monitor candidate quality and response, and watch for hidden groups that platform algorithms may overlook.
• Treat sourcing as a changing set of tasks, not a fixed professional identity; adapt skills to where organizations have unmet needs.

Presenters and sources

Presenters and participants: Hung Lee (host); Alla Pavlova; Elise (MetaView); Guillaume; Wim; Chloe Morrison; Marcel (surname unclear in the subtitles); and Chad Horn. The subtitles also thank Marie Vanameir, though her role in the discussion is unclear.

Other sources referenced: SOSU EU 2026; Glenn Cathy (sourcing definition and search practices); DHH, identified in the discussion as the creator of Ruby on Rails; and K.’s technical sourcing work on GitHub.