Prospect research
Reads LinkedIn profile, company site, recent news. Pulls one specific personalization angle per prospect (recent post, funding event, role change, mutual connection).
Sofia eats 8 hours of SDR work per day. She researches prospects, drafts personalized cold email in your voice, sequences follow-ups, monitors replies, and escalates hot leads with full context. You get back the 30 minutes of human judgment that actually moves the deal forward.
Sofia is an AI sales outreach agent that runs autonomously: she pulls target prospects from your CRM or list, researches each one across public sources (LinkedIn, company site, recent news, podcast appearances), drafts a personalized first-touch email in your voice, sequences multi-step follow-ups, monitors replies, triages responses (hot lead → human, polite no → close-loop, out-of-office → reschedule), and books meetings on your calendar when a lead says yes. Built on Claude Sonnet. Operational tier. 93% eval pass rate. Production-ready. Typical throughput: 50–100 personalized emails per day.
theagency47 · Updated May 2026| Job-to-be-done | Generate qualified meetings from a target list by handling the repetitive outreach steps autonomously |
| Tier | Operational (departmental flow) |
| Underlying model | Anthropic Claude Sonnet |
| Trigger | Daily scheduled run + event-driven on incoming replies |
| Inputs | Target list (CRM segment or CSV), brand voice document, sequence template, calendar availability |
| Tools available | Web search, LinkedIn (public), CRM read+write, Gmail send/read, calendar (Cal.com / Google Calendar), Slack notification |
| Autonomous decisions | Personalization angle per prospect, send timing within business hours, follow-up cadence, reply categorization |
| Escalation rules | Hot reply → human (full context attached) · Out-of-policy phrasing detected → human · 3+ bounces from a domain → pause and notify |
| KPIs measured | Reply rate, positive-reply rate, meetings booked, time to first response, opt-out rate, cost per meeting |
| Eval suite | 25 test cases (70% common, 20% edge, 10% adversarial). Pass rate 93% as of May 2026. |
| Audit trail | Every email logged with prompt version, research sources, and decision reasoning |
Reads LinkedIn profile, company site, recent news. Pulls one specific personalization angle per prospect (recent post, funding event, role change, mutual connection).
Drafts in your voice using a brand voice document trained on your highest-performing past emails. No "I hope this email finds you well." No bullet-pointed value props. Real human-sounding outreach.
Sends touch 1. Waits. Reads reply or silence. Sends touch 2 (different angle, not "checking in"). Sequence ends after 3 touches unless reply.
Reads each reply, categorizes (interested / not now / not interested / wrong person / out of office), routes accordingly. Hot leads get human attention same-day.
When prospect says yes, Sofia checks your calendar and proposes 3 specific times. Confirms the booking. Adds the meeting to your CRM with full context attached.
Updates contact records with new info learned during research. Logs every touch. Marks dead leads. Closes the loop on stale sequences.
Custom integrations with proprietary CRMs or sequencing platforms quoted as add-ons.
Sofia is an AI sales outreach agent. Researches prospects, drafts personalized cold email in your voice, sequences follow-ups, triages replies, books meetings. Operational tier. Built on Claude Sonnet. 93% eval pass rate.
Not as standalone SaaS. Sofia is a reference template we customize per client during a Spark or Workforce Starter engagement. Customization makes her speak in your voice and target your ICP.
Apollo/Outreach are sequencing platforms. Sofia is the agent that does the personalization work inside the sequence. She can run on top of Apollo, or standalone via Gmail + HubSpot. Sales-engagement tools template content; Sofia generates personalized content per prospect.
Sofia template customized to your business: deployable as Spark (€2,500, 14 days) for a single business or as part of a Workforce engagement. Operating cost €30–€150/month for Claude API usage at typical volume.
Not when configured correctly. Sofia respects rate limits, business hours, sender warm-up periods, and bounce thresholds. She also writes well, and "well-written email" is the single biggest deliverability factor.
Yes. Sofia respects do-not-contact lists, includes unsubscribe links, processes opt-outs immediately, and only contacts prospects with a legitimate business basis under GDPR Article 6(1)(f). DPA executed before deployment.
Production deployment includes a 14-day human-in-the-loop review window, every draft is reviewed before send. After 14 days, Sofia moves to autonomous mode with sampling-based audit. Out-of-policy phrasing detection runs on every draft.