Sam Woods

A SaaS founder I advise doing ~$20M ARR fired his entire SDR team for an autonomous inbound system. About 6 weeks later, he called me in a panic because his pipeline was breaking. He had fired the human team to cut CAC. The data showed what actually happened: - Inbound volume jumped roughly 4x - Calendars filled with unqualified clicks - AE win rates collapsed by about 60% - The entire pipeline collapsed When an agent floods calendars with unvetted leads, your most expensive reps spend roughly half their week babysitting bad calls instead of closing margin. Automating top of funnel without mapping unwritten qualification rules just burns sales capacity.

The mistake isn't hiring an agent to run a workflow. It's hiring an agent to run a workflow before you've written down how a human does it well. You can't encode judgment you haven't examined. Most agentic builds skip this step entirely. They wire up the tools, set the trigger, and watch it run. Then wonder why the outputs are technically correct and operationally useless. The debugging loop that follows costs three times what the build did. Write the manual version out first. Every decision point. Every filter. Every condition where a human would pause and check something. That document is your actual architecture. The code is just a translation of it.

A client who loves working with you can still be the wrong person to offer your next program to. A second purchase needs a second problem to solve, so the buy-more score looks for three things: - a result they'd name - engagement rising against their own normal - a new problem they've raised in their own words For a referral, swap the new problem for praise they gave without being asked. A referral puts their name next to yours, so they have to be pleased enough to vouch for you. Both readings need the result. A client who hasn't had anything from you yet won't buy more or send you a friend, however much they like you.

If you run agents in your sales or ops pipeline, give each one an operational boundary contract. It's one markdown file that lives in the agent's context, six lines per workflow: - Trigger: the exact event that wakes the agent up. - Input: what it is allowed to read (and what it cannot touch). - Action: the single discrete change it makes. - Gate: the validation rule it must pass before writing data. - Fallback: where it routes when confidence drops below 90%. - Log: the one-line summary it posts to Slack. Then add one standing constraint to the system prompt: "If an action requires guessing outside these six boundaries, stop execution and notify an operator." Most broken automations happen because teams hand an agent a goal instead of a fenced lane.

Defaulting every background task to frontier models is a quiet margin killer. We found roughly $2,600 a month vanishing across ~50,000 routine calls. Jesus take the wheel.

When I ask consultants and coaches how they decide which client hears about their next offer, the honest answer is nearly always whoever they spoke to last. The clients who are ready have usually said so already. One asks on a call whether you also work with their team. Another emails, without being asked, to say what changed for them since they started. None of it turns into a conversation, because the month fills up and nobody reads it back. So I'd hand that reading to an agent. Once a month it goes through every client's last ninety days of call notes, email and payments, and scores four signals, each with a dated quote or event behind it: - a result they'd name - engagement rising against their own normal - a new problem you could solve - praise nobody asked for Then it reads the score two ways, once for who's ready to buy more and once for who's ready to refer you. It holds back anyone with something open or anyone you asked recently, and cuts the list at the number of conversations you have room for this month. It never contacts a client or names a price, and for each client who makes the cut it writes you an opening question built from something they said. The first version takes about forty minutes: one sheet of clients, ninety days of notes, and one prompt in the AI you already pay for.

An agency owner I work with lost a ~$4,000/mo retainer to an onboarding delay: > Closed the deal > Welcome email delayed > Kickoff pushed > Client rescheduled > Internal prep meetings stacked up > First deliverable landed roughly 19 days in The client spent those 19 days wondering if they made a mistake. Delivering automated audit notes within about 90 minutes of payment flips the dynamic immediately. The client holds something tangible on day one before anyone even joins a kickoff call. Painful.

Had coffee with an agency founder doing roughly $1.1M who automated about 70% of his outbound research and lead scoring: - Research time dropped from 4 hours a day to ~20 minutes - Outbound pipeline volume doubled in 30 days - Close rate stayed completely flat He assumed more qualified leads would automatically solve his revenue target. An audit I ran showed something else: His delivery team was already at capacity, so new client start dates slipped by three weeks, and two prospects backed out before onboarding even started. Fixing front-of-funnel conversion without checking delivery bandwidth just builds a bigger bottleneck downstream.

A solo consultant I spoke with running a ~$180k practice across 6 retainers tracked their time for a month: He assumed client delivery was eating most of his week. The audit showed something else: - About 69 total operational hours logged - Roughly 53 of those hours spent shuffling data across ~11 browser tabs - Status updates, client reporting sheets, scheduling links, and invoice chasing Deep strategy barely moved the needle on his workload. When you operate solo, tab switching never feels like real work. It registers as two minutes to copy a metric, three minutes to update a client doc, or five minutes to send an invoice reminder. Across 6 retainers, that invisible coordination ate roughly 77% of his operational hours. He spent three quarters of his month acting as manual middleware between browser tabs. C'mon man.

Audited an agency losing a ~$4,000/mo retainer across ~19 days of onboarding: He assumed bad client fit. The data showed silence killed it: - First deliverable took ~19 days - Diagnostic notes sent in ~90 minutes stop the churn Latency destroys retainer margin every time.

11 of 14 operators I interviewed were deploying multi-agent swarms for tasks a basic Zapier webhook solves. The operational ladder: 1. Assisted (prompts) 2. Automated (webhooks) 3. Augmented (agent + human gate) 4. Autonomous Jumping straight to 4 builds brittle overhead.

Audited a ~$180k/mo DTC skincare brand fighting an ~14% return rate: The team rewrote product descriptions, swapped photos, and retrained customer support, but returns didn't stop. Raw ticket logs showed: - Buyers layered a serum and cream - Formula pilling triggered breakouts Fixing packaging instructions protected the margin.

If your Claude skills never trigger, check the description field. Audited 23 skills recently. 18 had never been invoked once. Same bug across all 18. The description read like a summary: "This skill helps create marketing content for B2B SaaS companies" The model picks a skill by matching a condition, and a summary gives it no condition to match. So the first line has to be the trigger: "Use when drafting marketing content for a B2B SaaS company" Five skills doing all the work while 18 sat in the folder.

When founders blame model limits for generic agent output, they usually have the diagnosis wrong: Upgrading from 3.5 to Opus does not solve the problem when an agent is starved of unwritten context. About 80% of what makes the business work lives inside the operator's head. The agent only reads the ~20% that is documented. Paying for a frontier model just runs the same missing context through a pricier engine.

Amateurs wrongly believed sales calls handle objections. The systems show dead pipeline is decided before booking: - Intake forms filter fit upstream - Agent decision logs isolate objections early If you have to explain delivery on a call, intake already failed.

Watched an operator build an agent for a job that runs roughly 8 times a year: - ~15 hours building triggers and watchers - About 20 minutes to do it manually If a workflow executes fewer than 10 times a year, keep it as an ad-hoc prompt or do it by hand.

You cannot write down formatting rules in a prompt. I tried for years. The instructions kept getting longer while the output stayed slightly off. What actually works is boring: paste in one finished artifact made by hand and tell the model to match that cadence.

An agency owner doing about $900K I audited had built up a sophisticated AI software stack: He believed the tools themselves were the leverage, even though none of them were tied to time, margin, or revenue. The audit data showed something else: The audit showed he was already sitting on the components of an owned media operation without capturing the margin: 1. Unfiltered teardowns: Client work already produces raw operational breakdowns that can be published almost as-is. 2. Zero production overhead: The material comes directly from delivery work, costing almost nothing in extra creation time. 3. Proof before the pitch: Showing real workflow math closes inbound interest before an offer is even mentioned. 4. Retainers over sponsors: Instead of selling brand deals, the content converted readers straight into ~$8,000 monthly retainers. Treating client delivery as an asset library cuts acquisition costs to roughly zero and protects service margin.

I want the ops setup that quietly protects pipeline conversion without asking for a Slack huddle. An agent handling intake, scoring friction points, and routing the prospect before a call hits the calendar. Your team stops acting as manual middleware between browser tabs, and client delivery actually gets protected.