Our LinkedIn has been bombing lately, and Chris isn't convinced another viral hit is the fix. His public pricing post went viral and he calls it a one-hit wonder. So we argue it out. Repeating the same core ideas until everybody gets them, controversial takes that still pull in the right people, the AI job Chris built to find industry posts worth commenting on, and small wins that compound over time.

Then lead gen. AI tools that find leads, call them and book the meeting, inbox pitches promising 50 leads a month, and why big promises fall flat without a believable mechanism behind them. Alex hangs up on every cold call, Chris says it depends on how aware the buyer already is, and we get into the two ways a pitch misses, all promise and no mechanism or all features and no benefit.

We also get into Chris prepping his first BrightonSEO talk and how new speakers start in smaller rooms, building a Google Knowledge Graph entity so Google and AI tools know who you are, how easy LLMs are to game, and GEO as a way for SEO agencies to raise prices. Plus customers sending AI-written support tickets, Chris hiring a running coach after two years of relying on AI while Alex sticks with his AI coach at day 66, and Opusfived, the parody where Claude can't change one button's color.

No Custom Work is Chris and Alex from Wayfront, the client portal software for agencies selling repeatable productized services. New episodes every Wednesday.

Try Wayfront free for 14 days, no credit card required: https://wayfront.com/trial

Your client asked for monthly updates. They have not opened one since March. Chris admits he does the same thing to the agencies he pays, which kicks off a conversation about what a client update actually has to look like if you want it read. TLDR at the top, the full detail one click away, a soft upsell at the end, and a straight answer when the numbers are bad instead of six months of "we're still ramping up."

The second half gets into the AI question every agency is working out. Clients run your feedback through AI and send it straight back to you. Some industries will publicly flame you for touching AI at all, while plenty of clients are already asking you to use it.

Also in this one: why agencies spend four months setting up a portal and never launch, connecting AI to a client portal through MCP, the weekly call where Chris and Alex audit their own product with fresh eyes, shipping a feature with no changelog entry, and whether Grok, Claude and ChatGPT are interchangeable now.

No Custom Work is hosted by Chris and Alex from Wayfront, the client portal software for agencies selling repeatable productized services. New episodes every Wednesday.

Try Wayfront free for 14 days, no credit card required: https://wayfront.com/trial

Chris opens with a hot take: agency owners need to stop chasing the SaaS dream. He keeps meeting people who tried building software, watched it flop, and want another go, when distribution was the constraint the whole time. With AI handling more of the delivery, he thinks the agency model scales better now than it used to.

From there: demand constrained vs fulfillment constrained agencies, why better margins let you outspend competitors on client acquisition, and how budget brackets mean a 5K client and a 50K client are being served by agencies that never actually compete.

We spend a while on white labeling. Alex would mark up 1.5x, the industry charges 3x, and the real value add turns out to be niche fit and curation. That leads into why adding services outside your core competence quietly degrades everything, and why doing one thing repeatedly compounds because you end up fixing the system instead of the project.

Then the AI half. What clients will still pay agencies for in ten years (risk and accountability, mostly), Chris's simplest AI workflow using MCPs pointed at a folder of plain text files, a check-in on Buddy, Grokbot and Hermes, and the CEO agent pattern people are running.

It closes on agentic commerce: a client's AI shortlisting three agencies and buying the service without a human clicking anything. Alex thinks it's too early, Chris wants to build it anyway, and Wayfront's central MCP endpoint is coming either way.

Every agency owner eventually thinks about building a SaaS. Chris is giving a BrightonSEO talk about exactly that this autumn, and his thesis is that people want the SaaS metrics without understanding the problems that come attached.

We disagree for most of the first ten minutes. Alex argues a cheap product is easier to sell than a $5K retainer. Chris argues the process is identical at any price, and that the real advantage of a high LTV client is the marketing budget it unlocks for you. Both of us end up moving.

Then it turns into a conversation about what AI actually changes. Chris calls it the AI dividend: delivery gets faster and cheaper, so you hand that back to clients as more value rather than keeping the margin. Someone emailed Wayfront recently asking whether the platform supports fully AI delivered services, and it does, so we go through the setup with the MCP, order forms, and an agent that fulfils the work and updates the order.

In this episode:

- What agency owners actually want from SaaS: MRR, margin, zero marginal cost per client

- Whether a $100 product is genuinely easier to sell than a $5K retainer

- Why a $3K and a $10K client need the same hand holding

- Jason Cohen's take on what "too expensive" really means when a client leaves

- How much of your churn is your fault, and how much is just the client's own situation

- The AI dividend and passing the savings on instead of pocketing them

- Running a fully AI delivered service with the Wayfront MCP, order forms and agents

- The dumb AI workflow that works: tell your agent to check the MCP every 15 minutes

- Press release distribution as a service that is already automated

- wayfront.ai and everyone shipping their own AI harness

- ChatGPT voice mode pulling status updates from your other chats

- Testing your own signup flow to refine it, plus swapping magic links for OTP codes

- Launching a client portal in a market this crowded

Chris pitches an AI sales agent for agencies. A week after delivery it reads the thread, sees the client was happy, and drafts a review request you approve with one click. Same when an old client logs back into the portal, and when a card fails: a personal founder email instead of the automated one nobody opens.

From there: hourly billing caps your income now that AI multiplies output, Jason Cohen's Hidden Multipliers on spotting unprofitable clients, and why SOPs clear enough for anyone to follow can now be automated.

Also in this one: SparkToro's report that marketing agencies are 500% more likely to use Claude, Chris's move to Grok, LinkedIn's new "AI slop" flag, and the reactivation email so plainly written a customer thought it was phishing.

Does cold outreach actually work? We get into the honest version this week: relevance and timing beat raw volume, the pros run hundreds of inboxes, and short emails win because nobody reads anything. We also revisit our physical cold mail experiment, where we sent packages to agencies and got exactly one call back.

From there it's founder brands on LinkedIn: why walls of text flop, why hooks and memes win the feed, and what positioning looks like for a personal brand. Plus the $5 Fiverr CEO verdict and the shady world of pay-to-publish directories.

Then the practical stuff: the Wayfront redesign is rolling out to every workspace with a mobile experience that finally feels like an app, the feedback button is back, and Chris finished a BigQuery data warehouse that pulls ads, analytics, and survey data into one place so AI can surface insights instead of static dashboards. And yes, our AI agent quietly changed the pricing on a side project, and someone tried to pay. We wrap on why leads need 7 to 10 touchpoints before they buy.

Last week we said the native apps had caught up and Buddy could go. One week later the ChatGPT app is lagging on websockets, Chris is screen sharing into the Mac Mini to get anything done, and Buddy is looking less retired than advertised. That leads into the agent harness market, the $99/mo wrappers people are paying for, and why one tool can never fit everyone.

Then we get into Stripe. It used to be the design gold standard and now the proration setting stays hidden until after you click continue, which is a strange place to put the thing that decides when your client gets charged. Same lesson applies to us: our own CLI and MCP sat under Modules for months and customers had no idea they existed.

We also cover the Airtable sale to Bending Spoons and what liquidation preferences leave for founders, the illusion of choice across review sites that quietly own each other, why we pulled the client logo strip off our website in favor of case study numbers, and why "book a call" is fading as a CTA now that buyers show up already knowing exactly what they want.

Plus a Backlink Pilot update and where the Wayfront redesign stands, including native dark mode.

Buddy is going to the pet cemetery. One week after planning a course on building your own AI agent, Chris and Alex are retiring theirs: the native ChatGPT and Claude apps finally caught up, Buddy was 50% slower for coding, and his support answers were sauce. They call it the "they built the app" trap, where you invest in workarounds right before the official version gets good.

From there it's why you can't build a product for everyone. Refusing to ship proposals, sacrifice as proof of positioning, and how agencies end up specialized by cloning their one great client.

Alex also breaks down his AI weight loss setup: a Claude project that tracks every bite using memory files and his Notion recipes, roughly 4kg down in a few weeks (and yes, it scolds him about protein). Plus Mac Minis, local LLMs, and running an entire company from a phone.

They close with the plan to revive Backlink Pilot as an AI slop playground: a zero-dollar business running fully automated content, why "fully automatic" really means 200 hours of setup, and AI avatars that now pass job interviews. Next week: hiring a CEO on Fiverr to run the slop company.

Two episodes ago we said agencies should put their prices on their website. We turned it into a LinkedIn post and it did 230,000 impressions, 476 likes and 120 comments, most of them people arguing with each other.

So this episode we go back through the counterarguments properly. What you do if your services genuinely range from $50 to $50,000. Whether price works as a quality signal in services nobody can touch before they buy. When "starts at" is the honest answer. And the comment that got us, which is that our own Plus plan has no price listed anywhere.

From there we get into the client portal redesign we're shipping, why every design goes stale about a year after launch, and how Apple quietly decides what counts as outdated for the rest of us.

Chris also updates us on personally texting trial signups, our AI survey showing almost no agencies are running real agents yet, and what an agent could actually do inside an agency.

We finish on the free plan idea: no fixed trial length, free until you land your first client. Whether a 14-day countdown creates any real urgency, and why we still won't take a credit card upfront.

Chris returns from an SEO conference in Estonia with the episode's headline story: court documents confirm Google uses Chrome and Analytics behavior data as ranking signals, and black hat SEOs exploited it by spamming complaint forms across random websites, since a completed form reads as satisfied user intent.

From there, we cover whether free tools still work as an SEO play now that everyone just asks AI, why AI skills break a week after you build them (and the multi-model "counselor" approach that fixed one), agencies repackaging the same services as generative engine optimization at premium prices, and how 2006-era spam tactics work again on AI search engines.

We close with what the rebrand to Wayfront taught us about brand SEO, the rise of risk-reversal offers, and productizing your taste and experience as AI skills rather than just your deliverables.