Self Service B2B Is A Lie

A large government buyer used a online ordering platform for standing orders.

Recurring orders.

Same products.

Same schedule.

The kind of feature that sounds so automated you'd put it on a slide deck.

I know I did, it was built just to get the contract.

But then they sign the dotted line, and start using the portal.

What the absolute fuck, they change their standing orders multiple times a day.

Every change generated a customer service ticket.

A human updated each one manually.

The customer thought they had self-service.

What they actually had was a request form with extra steps.

The system cost more to "automate" than the manual process it replaced. Because at least when they used to call, the rep could push back, batch the changes, or say "mate, you changed this an hour ago."

The portal made every micro-change feel free and instant.

It wasn't.

It cost the business time, accuracy, and margin.

Nobody budgeted for "standing order exception handling" because the feature was supposed to eliminate exceptions.

It created them.

This isn't a broken edge case.

This is how most B2B "self-service" actually works.

The real story

Self service in B2B is almost never self service. It's a portal skin stretched over manual operations.

The customer gets a login. Behind the login, someone is setting up their account, mapping their products, configuring their pricing, managing their credit terms, and fixing every exception the system can't handle.

The portal didn't remove the manual work.

It hid it.

And hiding it made it harder to staff, harder to cost, and harder to fix.

This is true across industries, across countries, and across company sizes. And it creates both a massive hidden cost and, for anyone paying attention, one of the biggest opportunities in B2B.

My first online order was a phone call

I started working in bike shops at 14. Branch transfers were just how it worked. Customer wants something you don't have, you call another store.

That part I understood.

At 19 I got a job working on the ecommerce side of the business, tech mogual, next zucc I though to myself as I stepped in the building.

This is where I learned what "self-service" actually looks like from behind the screen.

An order comes in online.

I pull it up in the ERP.

If the stock is in the warehouse, great.

It ships.

But about half the time, it isn't.

The warehouse doesn't have it.

A store does.

Maybe.

So I call the store. The store manager picks up and says "mate, I literally just built that bike yesterday. It's on the floor." Now I'm asking him to pull a built bike off the display, disassemble it, box it up, and ship it to a customer he'll never meet, loose out on KPI’s, so that the website gets a sale instead of his branch.

Sometimes they do it.

Sometimes they don't.

Sometimes the stock system says they have it but they sold it yesterday and haven't updated.

So I call another store.

Same conversation.

Same friction.

Same result half the time.

Then I go back to the customer and say sorry, we can't actually fulfil this. Lost sale. Bad experience. The customer has no idea that behind the "Add to Cart" button was a 19 year old on the phone trying to negotiate stock out of a branch manager who doesn't want to give it up for the last 3 hours.

This business I worked for went under.

There were a lot of reasons.

But this was part of it.

The ecommerce operation looked like a website.

Underneath, it was a negotiation.

Every order that wasn't in the warehouse was a manual, human, phone based, cross your fingers process with no guaranteed outcome.

Branch transfers are still the norm across NZ/AU retail.

But the ecommerce version of this problem is worse, because the customer expects the transaction to be instant and automated. They clicked a button. They paid. They think it's done. It isn't even close to done. It's just begun, and the first step is a phone call to someone who might not pick up.

That's why so many NZ retailers are terrible at online fulfilment.

The infrastructure isn't automated, it's a person.

And when that person is sick, or busy, or loses the argument with the branch manager? The order just doesn't happen. Nobody tracks the failure rate because nobody tracks the process.

Many Kiwi’s, and even Auzzie’s don’t understand why ‘my order arrived from temu before it arrived from my local dealer’.

Welcome to the mess behind the curtain.

The standing order that wasn't standing

Back to the government health buyer.

Standing orders are supposed to be the simplest possible B2B feature. Recurring order. Same products. Same schedule. Set and forget.

Except this customer didn't forget. They remembered constantly. Multiple times a day, in fact. Add this product. Remove that one. Change the quantity. Shift the delivery day. Add it back.

Each change went to CS.

CS logged a ticket.

CS updated the order manually.

CS confirmed the change.

Repeat.

The volume overwhelmed the team. What should have been a background automation turned into a full time support function that nobody anticipated and nobody owned.

Here's the thing most people miss: the portal made this worse, not better. When this customer used to call, the rep could say "I'll batch those changes at 3pm" or "are you sure? You changed this twice already today." The phone call had a natural friction that protected the business.

The portal removed that friction.

Every click felt instant and free to the customer. Behind the portal, nothing was instant and nothing was free. Someone still had to process every change. But now there was no pushback mechanism. No batching. No "are you sure." Just an open pipe from the customer's whims to the operations team's workload.

The "self service" portal that was promised on the deck created more work than the phone calls it replaced. Nobody budgeted for this because the business case assumed the feature would eliminate manual work. Instead, it generated a new category of manual work that was harder to track and impossible to push back on.

Your account takes 3 days because 4 departments touch it

In consumer ecommerce, you sign up in 30 seconds. Email, password, done. You're shopping.

In B2B, before a customer can place a single order, the following has to happen.

Account created in ERP.

Credit application reviewed by finance.

Customer mapped to a sales rep or branch.

Product visibility rules configured.

Customer-specific pricing loaded (because in B2B, almost nobody pays the same price).

Delivery logic set up: cutoffs, routes, exclusions, minimum orders.

User permissions assigned.

Often, partner specific mappings for EDI or integration.

Tax configuration based on entity type and location.

That's four departments minimum.

Finance. Ops. Sales. IT.

Each with their own queue, their own priorities, and their own definition of "done."

Average time from "I want to order" to "I can order" in mid-market B2B? Days. Often weeks.

Sometimes longer if the credit check gets stuck or the pricing matrix needs manual setup.

The customer sees a sign up form. Behind it sits an operations pipeline that nobody owns end to end. Sales hands it to finance. Finance hands it to ops. Ops waits on IT. IT says "we'll get to it."

And the gap between "account created" and "account actually working" is where trust dies. Your new customer is sitting there with a login that doesn't do anything yet, wondering why they bothered signing up when they could have just called a rep and placed an order.

Most B2B businesses don't measure this gap.

They should.

It's the first impression of your "self-service" experience, and for many customers, it's the last.

The iceberg behind the login

Zoom out and look at what actually has to function before a B2B customer can "self-serve."

Account setup.

Customer mapping.

User permissions.

Product visibility rules.

Customer specific pricing.

Credit terms.

Delivery logic.

Rep assignments.

Tax configuration.

Payment terms.

Order approval workflows.

Invoice history linkage.

None of this is in the portal. All of it is done by a person before the portal works. And when it breaks?

A person fixes it.

When the price is wrong?

A person corrects it.

When the delivery window doesn't match the customer's expectation?

A person calls.

The portal is the tip of the iceberg. The iceberg is operations.

Most B2B businesses spend 80% of their digital budget on the tip. The slick UI. The responsive design. The product search. The checkout flow. All the stuff that shows up in a demo and impresses the board.

Meanwhile, the thing that determines whether the customer can actually order, and whether the order is actually right, and whether the delivery actually arrives when promised?

That's held together by people, spreadsheets, and a healthy dose of hope.

You're not alone

This isn't just a B2B distribution problem.

This is a pattern everywhere self service touches complexity.

Let’s go back to a story i’m sure you know.

Builder.ai raised $445 million selling "AI-powered" app development.

Their AI assistant "Natasha" was supposed to build apps like ordering a pizza. That was the marketing line.

The reality?

700 engineers in India doing the work manually.

Customers thought they had self-service automation. They had an outsourced dev shop with a chatbot stapled to the front.

It gets worse. Revenue was inflated 4x through fake invoicing between Builder.ai and an Indian partner company. The Wall Street Journal flagged the human labour issue back in 2019. Nobody cared because the story was too good.

Microsoft invested.

SoftBank invested.

Qatar's sovereign wealth fund invested.

The company hit a $1.5 billion valuation.

Then the audits happened. $220 million in claimed 2024 revenue turned out to be closer to $55 million. The company went bankrupt in 2025. Prosecutors got involved.

Same pattern.

Slick frontend.

Manual backend.

But in this case, unsustainable economics.

Amazon's "Just Walk Out" stores were supposed to be pure computer vision. Walk in, grab your stuff, walk out. AI handles the rest.

In 2022, 700 out of every 1,000 transactions were manually verified by a team of over 1,000 workers in India. Customers walked out thinking they'd just experienced the future. A human watched them shop and tagged every item. Receipts arrived hours late because someone had to review the footage. Amazon quietly killed the program in its grocery stores.

Expensify's "SmartScan" receipt feature was marketed as AI reading your receipts. When the AI failed, the system farmed transactions to Amazon Mechanical Turk workers. Real humans viewing full credit card numbers, names, and addresses.

X.ai marketed "Amy," an AI scheduling assistant. Behind every email was a human. One 24-year-old told Bloomberg they were manually scheduling meetings, ordering food, and managing calendars while pretending to be software.

Waymo sells itself as a fully autonomous, 24/7 robotaxi. The cars may be driverless, but the operation still has around 70 remote assistance agents on duty at a time, possibly more they don’t admit too, stepping in to provide “limited guidance” when the system gets stuck.

Presto Voice sold "AI drive through ordering" to Carl's Jr and Del Taco. The AI was humans listening to orders and placing them manually.

Every single one of these followed the same playbook.

Sell the automation.

Hide the manual layer.

Hope you can close the gap before anyone notices. Most don't close the gap. The manual layer becomes the product. And then the economics collapse.

Why it persists

The customer doesn't care how it works, they care that it works.

This removes market pressure to actually automate.

And the manual cost is distributed, account setup in ops, pricing in finance, sync failures in IT. Nobody sees the total because nobody owns it.

Meanwhile, businesses launch the portal first and promise to automate later, but 'later' never comes because the manual workarounds become infrastructure until you add it up.

And almost nobody adds it up.

Building real automation is expensive and slow.

So businesses launch the portal first and promise to automate later. This makes sense in theory.

Ship the MVP.

Fix it in production.

Except "later" never comes because the manual workarounds become embedded in the operating model. People get hired to do the manual work. Processes form around it. Budgets absorb it. By year two, the manual layer isn't a stopgap. It's infrastructure.

But the honest truth is that the manual layer is often better.

A human who knows the customer's account, knows their history, and can make a judgment call on a substitution or a credit issue is sometimes worth more than an automated system that follows rules. The problem isn't that humans are involved.

The problem is that nobody planned for it, nobody costed it, and nobody knows how much it's actually eating.

The opportunity in the open

The businesses that actually automate the manual layer have a structural cost advantage.

Not a UX advantage.

A margin advantage.

The opportunity sits in four places.

Onboarding speed.

If your competitors take 5 days to get a new account ordering, and you take 5 hours, you win the account before they've finished their credit check.

The business that cracks self serve account creation in B2B, with real credit decisioning, real product mapping, and real delivery configuration, doesn't just improve UX.

It compresses the sales cycle.

Every day between "I want to buy" and "I can buy" is a day the customer might call someone else. Some businesses have slick on boarding, but in practice, setup always fails leaving you with credit checks not worth the paper they are written on.

Exception handling as a product.

Standing orders that actually let customers self manage changes without generating CS tickets.

Product substitution logic that works without a human.

Delivery scheduling that accounts for cutoffs, holidays, and route constraints without someone in ops manually overriding the system.

Every exception you automate is a headcount you don't need and a customer you don't lose. Most B2B businesses treat exceptions as annoyances, smart ones treat them as product features.

Stock truth and fulfilment routing.

Remember the Bike shop? Every order that wasn't in the warehouse became a phone call, a negotiation, and often a lost sale.

The retailer that solves real time, cross location stock visibility and automated fulfilment routing doesn't just improve the customer experience.

It captures revenue that currently evaporates.

The flagship store model makes this worse, because flagship locations hold the most stock but have the least incentive to ship it out.

Their KPIs are foot traffic and in store sales, not ecommerce fulfilment.

So your best inventory sits in the location least motivated to release it.

The fix isn't just software. It's incentive design.

But you can't redesign incentives until you can see the cost of the current model: every declined transfer, every cancelled order, every customer who bought from someone else because your website said "in stock" and your store said ‘nah.’

Visibility into the real cost.

Most B2B businesses have no idea what self service actually costs them.

They know the portal cost.

They know the hosting cost.

They don't know the cost of the 15 people across four departments who keep the portal from embarrassing the business.

The company that instruments this, that actually measures cost per account setup, cost per exception, cost per standing order change, finds margin in places nobody else is looking.

You can't automate what you can't measure.

And you can't measure what you don't acknowledge exists.

This is also where the biggest consulting and SaaS opportunities live.

Tooling to automate B2B onboarding, exception handling, and operational workflows barely exists in the mid market.

Enterprise has Salesforce and SAP. Small business has Shopify. The $10M to $200M B2B distributor?

They have a custom portal and a prayer. That gap is enormous. Whoever fills it without ductape will own the most valuable layer of B2B commerce: the plumbing.

When it all comes crashing down

You can't automate self service if your ERP is the source of truth and your ERP takes 48 hours to sync a new customer.

The portal is downstream.

If the upstream systems are manual, broken, or slow, the portal just becomes a faster way to generate tickets.

I've seen businesses spend $500K on a new ecommerce frontend while their product data syncs once a day and their pricing updates require a CSV upload.

That's not digital transformation.

That's digital decoration.

But truthfully, lot of B2B buyers don't actually want full self-service.

They want a rep who knows their account, knows their history, and can make a call on a substitution or a credit issue.

Full self service without an escape hatch to a human is how you lose accounts to a competitor whose rep picks up the phone.

Pricing updates don't need a human. Account setup validation doesn't need a human. Standing order modifications don't need a human. But a customer asking whether they should switch from a 5kg to a 10kg carton because their volumes changed?

That's a conversation.

The goal isn't to eliminate humans. It's to eliminate the manual work that doesn't need them. Which is why every B2B portal demo looks great while every production deployment is chaos.

The first order works.

The catalogue loads.

The account is pre-configured.

Applause all round.

Then a real customer logs in with a multi-location account, customer-specific pricing that's 6 months out of date, a delivery address that doesn't match their ERP record, and a product range that's missing 40% of what they actually buy.

The demo didn't break because the demo didn't have exceptions.

Production is nothing but exceptions.

If your testing didn't include a customer with 14 delivery locations, a mix of weekly and fortnightly orders, and a product catalogue that was last updated in November, you didn't test.

You performed.

This is the Builder.ai trap.

If you know the manual work exists and you hide it, you're building a business on a lie.

If you know it exists and you cost it, you're building a roadmap. The difference between a company that scales and a company that implodes is whether the manual layer is a secret or a strategy.

Every business has manual work behind its portal. That's not the problem. The problem is pretending it's not there, because that's how you end up with 700 engineers in India and a federal investigation.

So What.

If you're building or running B2B self service, cost the manual layer first.

Not the portal.

Not the UX.

The people.

How many account setups per week? How many exception tickets? How many standing order changes? How many "quick fixes" that take 20 minutes each? How many phone calls to check stock at another location? How many pricing corrections because the sync didn't run?

That's your real cost of self-service. The portal is the easy part. The operations behind it are the product.

The businesses that win in B2B aren't the ones with the best portal. They're the ones who figured out what to automate behind the login, and what to leave with a human who actually adds value.

Self service means the customer serves themselves.

If your team is still doing the work, you just built a nicer way to ask for help.

Shout out to all the legends I've ripped info from for this piece:

The Information // Bloomberg // The Wall Street Journal // Rest of World // The Pragmatic Engineer // Quartz // Gizmodo // Business Standard