A written SOP does not run itself

We spent close to two hours with someone who has run stores and sales organisations for 18 years at a large home interior company in Korea. One line stayed with us longer than the rest.

"I don't believe the close rate the marketing team shows me. That number was already massaged once on its way up from sales."

At first it sounded like a data quality complaint. It was not.

In the field, the work is already decided in messages, on calls, in showroom consultations and site visits. Then someone has to summarise all of that and type it into the company system a second time. A one hour consultation shrinks to two lines, and the reason a customer hesitated becomes a status value called "reviewing". Those two lines pass through a branch manager on the way to head office, and the number gets tidied up one more time. What he distrusted was not the marketing team. It was this pipeline.

They bought a CRM, and the dealers had no time to type into it

This organisation had rolled out Salesforce. Real usage across the franchise dealers never got past 30 percent, and at some point the sales manager's own KPI stopped being sales performance and became program adoption. The sentence a manager had to repeat on every dealer visit changed into this:

"Could you please enter this?"

The software was not bad. A dealer owner takes two customer consultations in the morning, goes out to measure a site in the afternoon, and builds a quote in the evening. There are no spare 20 minutes in there to open a laptop and retype what already happened. And if there were, those 20 minutes are worth more spent booking the next consultation. Field systems get abandoned less often because a feature is missing, and more often because they ask people to do finished work a second time.

This is not just one dealer’s situation. In Salesforce’s February 2026 survey of 4,050 sales professionals, reps spent 40 percent of their time actually selling. Data entry takes its share out of the rest.

What actually worked was a very plain rule

The same person described what worked when he ran the stores on Jeju Island. The rule was two lines long.

• Follow up call within one hour of the consultation ending
• Quote sent by the end of the next day

There was no special software. What there was instead was a clear "by when". A customer who walks out of your showroom visits three more companies that same evening. Whether your call reaches them in an hour or in two days decides whether your name is still in the running. Interior sales looks like a business of relationships and taste, but this is where the operating difference actually comes from.

The one hour mark has evidence behind it. In research Harvard Business Review published in 2011, firms that contacted a prospect within an hour of the enquiry were about seven times more likely to qualify the lead than firms that waited just one hour longer.

The trouble starts with scale. With 50 dealers and thousands of consultations, no human can keep count of who called within the hour and who sent the quote late. To find out, head office phones the dealers and asks, and the answer that comes back is a number that has been massaged one more time. So even a well written SOP stays in the document, and execution varies by person.

So we decided not to ask anyone to type it twice

We built Markhub on one rule. Nobody writes it again in a new system. The work starts in the conversation where it was decided.

We take it from the conversation. Sales consultations, site calls, meetings, team chat. We look first at the places where the work actually happens, and we go by what was really said rather than by what somebody summarised afterwards.

Promises become tickets. When someone says "I'll get you the quote by tomorrow", a task appears with an owner and a deadline on it. "Make the countertop grey instead of white" stays as a spec change. "We'll measure the site next week" becomes a date. There is no step where a person types it again.

AI surfaces what is running late. Once the ticket exists, the rest is automation. If 50 minutes have passed since the consultation ended and there is no record of a follow up call, it goes to the owner, and if it still isn't there, to the branch manager. If the customer said grey and the purchase order says white, it gets caught before the order ships. Which promises slip repeatedly, and at which dealer, accumulates on its own.

The screen a manager opens in the morning changes. Not an adoption table for 50 dealers, but the seven things that missed their deadline yesterday.

This is not an interior industry problem

In manufacturing sales, one phone call changes a unit price and pulls in a delivery date. The rep's notebook has the new terms, the system still has the original price, and the month end reconciliation does not match. In distribution, an order received over a messaging app, the purchase order and the shipping record disagree with each other, and nobody lines up all three until after something goes wrong.

It is the same shape. People already decide the work in conversation, and then the system asks them to enter it again. We are trying to remove that middle step, so that what was decided in the conversation becomes the work itself, and AI keeps checking whether that work actually got done.

How far do the numbers move

Everything above is structure. How much it actually changes has to be shown in numbers. Below are the targets we are validating in a POC that is running now. The first figure in each line comes from interviews and field observation. The second is a hypothesis, not a measured result, and we would rather say that up front.

• System adoption, meaning entry coverage. Under 30 percent today, 90 percent or more as the target, because the records are created from the conversation with no separate entry step.
• Follow up calls made within the hour. 40 to 50 percent today, 85 percent or more, with a deadline timer and automatic escalation.
• Quotes sent by the next day. Around 60 percent today, 90 percent or more, with deadlines assigned to the ticket automatically.
• Consultation to contract conversion. 10 to 15 points above the baseline, because response speed is the first variable in win rate.
• Wrong orders and spec mismatches. N per month today, 70 to 80 percent fewer, by matching the conversation against the contract and the purchase order automatically.
• Sales manager admin time. Over 10 hours a week today, under 3 hours, with roll ups and chasing automated.
• Reporting lead time. Weekly and massaged today, real time and from the source.

Converted into money, it looks like this. In an organisation with 10 sales managers, removing 7 hours a week of admin per person moves 3,500 hours a year back into selling. Add the average loss on a single wrong order, which is remaking material, delaying the schedule and handling the claim, at a few thousand dollars. A 70 percent reduction alone puts you past the subscription cost. This is not an argument for cutting headcount. It is an argument for changing what the same headcount produces.

What is still open

Three things are what we are actually measuring right now. How much work gets captured with no separate entry, how much follow up happens inside the promised window, and how much manager verification time drops. The rest of the list only moves after those three do.

The effect will differ by industry. But the more interviews we do, the clearer one thing gets. The field is not short on data. There are just three days between the moment a consultation ends and the moment it gets written into the system.

The same goes for an SOP. It starts working not when it is written in a document, but when the work is created from the real conversation, a deadline is attached to it, and someone verifies that it happened. The gap we are trying to close is the one in between.

Written from a field interview on 29 September 2026 with 18 years in interior distribution sales management.