I've been in enterprise sales for years. I've used Salesforce, Outreach, LinkedIn Sales Navigator, ZoomInfo, Gong, and every other tool the stack tells you to buy. In 2026, I'm replacing most of them — not with SaaS products, but with AI agents I built myself using Claude Code.
This is not a hot take. It's a build log. Here's what I've actually replaced, what I've augmented, and what still needs a human.
Context: I run enterprise deals at Freshworks — $200K–$2M+ ACV, 6–18 month cycles, 5–12 stakeholder deals. The tools I'm replacing are in that context. Your mileage varies for SMB or transactional sales.
What I've Actually Replaced
I run a multi-agent pipeline (3 parallel Claude agents) that researches any prospect in under 3 minutes: ITSM tools in use, recent IT incidents, decision maker backgrounds, ServiceNow contract signals, and a SPIN script tailored to that account. ZoomInfo gave me contact data. This gives me a sales brief.
✓ ReplacedI built a subject line scorer that runs via terminal and grades any subject line on 6 dimensions (curiosity, specificity, spam risk, CTA clarity, tone). It also rewrites the subject if the score is below 70. Took one evening to build, saved a monthly SaaS subscription.
✓ ReplacedApify + Claude Code scrapes the right contacts, discovers email formats, and drafts personalised outreach — all from a single /job-prospect command. What used to take 2 hours of manual LinkedIn work now takes 4 minutes.
✓ ReplacedWhat I've Augmented (Not Replaced)
I run /deal-review before every pipeline call. It scores deal health, surfaces MEDDIC gaps, and gives me the 3 questions I need to ask. But I still have to ask them. The AI can't read the room or handle the emotional dynamics of a stalled deal — yet.
→ Augmented/prep gives me a researched brief in 60 seconds: 3 tailored opening questions, ServiceNow displacement angles, the right close for the meeting, and a follow-up subject line. The call itself is still mine. But I show up sharper.
→ AugmentedWhat Still Needs a Human
No AI closes a CIO relationship. The trust, the reading of political dynamics inside an account, knowing when to push and when to wait — that's still entirely human. AI helps me prep for the conversation. It can't have it.
Human requiredAI will tell you a deal is at risk based on signals. It won't tell you that the champion just lost political capital in a reorg, or that the CFO is about to go on leave and nothing will get signed for 6 weeks. That intuition comes from years of pattern matching, not data.
Human requiredThe Pipeline I'm Building Now
The next level is a fully automated outreach pipeline I've been building in LangGraph. It pulls leads from a CRM, runs a 5-agent research chain against each one (LinkedIn, company website, recent news, ITSM signals, decision maker background), scores them for Freshservice fit on a 10-point scale, and for any lead scoring 7+, generates three things automatically: a personalised outreach report in Google Docs, a cold email draft directly in Gmail, and a SPIN-based discovery call script.
The scoring criteria aren't generic. I trained the scoring prompt on the specific signals that matter for enterprise ITSM displacement: is the company on ServiceNow (and how long?), have they had public IT incidents in the last 12 months, is the CIO or IT Director new in the last 18 months (new leadership = new vendor evaluation), and is there a recent funding event that triggers infrastructure investment?
The insight: A new CIO who inherited ServiceNow 6 months ago is not the same prospect as a CIO who bought ServiceNow and runs it as a matter of personal pride. AI can now tell the difference — if you give it the right signals to look for.
I'll publish the full build — including the LangGraph architecture and the scoring prompts — once I've run it against a live lead list and have real conversion data. That post is next.
The Honest Assessment
I'm not replacing my sales career with AI. I'm removing the administrative tax that eats 40% of a rep's week — manual research, writing and rewriting emails, building decks, formatting call notes — and redirecting that time to the things only humans can do: building trust, navigating politics, and closing.
The reps who will struggle in the next 3 years aren't the ones AI replaces. They're the ones who refuse to use it and get outgunned by the reps who do.
Every tool I've mentioned here took one evening to build. None of them cost a SaaS subscription. All of them are running right now, in a terminal, on my Mac Mini.
If you're an enterprise AE and want to build any of these — subscribe below. I'm writing up each one with the exact code and setup.
