Skip to content

Discovery Sprint

The Discovery Sprint is stage one: two of our forward-deployed engineers on site for two weeks, for a fixed fee. This page gives the plan day by day, describes the four deliverables you keep, lists what to have ready on day one and explains the readout.

This is the default plan. Before day one we move days to fit your shifts, your month-end close and the people who need to be in the room. With two engineers, two desks are covered at the same time.

Week one is for seeing the work as it is.

Day What our engineers do Who they need
1 Take your safety orientation, then hold a kickoff with the sponsor. Walk the floor and the front office, following one order from the customer’s purchase order to the shipment and the invoice. Sponsor, safety contact, someone to guide the walk
2 Shadow the order desk and customer service: orders by email, EDI and phone, then the order-status questions that follow. The people at those desks
3 Shadow purchasing and accounts payable: purchase orders, supplier follow-up, invoice matching. Buyers, accounts payable
4 Shadow scheduling. In a plant that is the scheduling board. At a distributor it is replenishment, receiving and returns. The scheduler or the warehouse lead
5 List every system and spreadsheet with the IT manager or ERP administrator. Confirm that read access to the ERP data works. Walk the draft Operations Map back to the people they shadowed. IT manager or ERP administrator, the people shadowed

Week two is for measuring, ranking and building the prototype.

Day What our engineers do Who they need
6 Measure where the hours go: count orders, purchase orders, invoices and emails in the ERP data and the shared inboxes, then set that against the handling time they saw at each desk. Short follow-up questions at each desk
7 Rank the workflows and review the draft Automation Backlog with the sponsor. Agree on the workflow at the top. Start the Working Prototype, on the AI model provider and in the place agreed in writing before day one. Sponsor
8 Build the Working Prototype on your real data. The person who does that work today
9 Run the prototype on real work beside the person who does it today, then fix what they find. Finish the Operations Map, the Systems and Data Audit and the Automation Backlog. The same person
10 Present the findings at the readout. Sponsor, desk managers, IT manager or ERP administrator

Shadowing means an engineer sits beside the person doing the work and watches the real thing: the purchase order PDF that lands in the customer service inbox, the screen it is keyed into, the spreadsheet that gets updated afterward. The engineer asks why at each step. They measure the work, not the person.

You keep all four whether or not you continue.

How the work actually flows, step by step, including the spreadsheets, inboxes and workarounds. For each workflow it shows:

  • What starts it: an emailed purchase order, an EDI file, a phone call.
  • Each step in order, with the role that does it and the system, spreadsheet or inbox it happens in.
  • What gets keyed twice, where work waits and where it goes wrong.

Every system involved, what data it holds, how clean it is and how to get data in and out. There is one entry for the ERP, one for each other core system and one for each spreadsheet or shared inbox the work depends on. Each entry records who looks after it, what it holds, which ways of attaching it supports and what is wrong with the data, such as duplicate customer records or prices kept in a spreadsheet beside the ERP price list. For the ERP, it also records which screen-level checks an interface does not run. Attaching to your existing ERP explains the ways of attaching.

The workflows, ranked by hours spent, cost of errors and how feasible they are to automate. Each entry gives the hours measured in week two, what an error costs in plain terms (a wrong price means a credit and a rebill) and what stands in the way: data that is hard to reach, steps that are not fully digital, many kinds of exception. The Embedded Build works from the top of this list.

The top workflow from the backlog, running on your real data. If that workflow is sales order entry, the prototype reads real emailed purchase orders, matches the part numbers and prices against your ERP data and shows a draft order beside the original for your order desk to check.

It is a prototype. It reads your data and writes nothing to your ERP, because the sprint runs on read access. It shows what the workflow can look like on your data. It is not yet software to run a desk on.

  1. A sponsor with time on day 1, day 7 and day 10.
  2. A shadowing schedule. Named people at the order desk, in purchasing, in accounts payable, in scheduling and in customer service, each told in advance who is coming and why.
  3. Read access to the ERP data, requested from your IT team before day one, with a named account for each engineer. It has to work by day 5, because week two counts from that data.
  4. Real documents. Access to the shared inboxes and spreadsheets the work runs through, or copies of recent documents from them, so the prototype runs on real work.
  5. A place to sit in or next to the front office.
  6. A safety orientation booked for the first morning, with a list of the protective equipment each area requires.
  7. A written agreement on AI and data, made with your IT manager before day one: which AI model provider the Working Prototype may use and on what terms, and where our engineers’ working copies and the prototype run. No document of yours reaches an AI model before that.

What we need from you has the list for the whole engagement.

The readout is on day 10. The sponsor attends, with the managers of the desks that were shadowed and the IT manager or ERP administrator. The people who were shadowed are welcome. Our engineers go through five things, in this order:

  1. The Operations Map, one workflow at a time. The people who do the work say where it is wrong.
  2. The Systems and Data Audit: how clean the data is, which of it can be reached and how.
  3. The Automation Backlog: the ranking and the reason for each position.
  4. The Working Prototype, run live on real documents from that week.
  5. The proposal for an Embedded Build: the workflows it would cover and the level each is planned to reach, a proposed workflow owner for each, the on-site schedule and the number of engineers and months.

You leave with all four deliverables. You decide whether to continue, and you do not have to decide in the room. If you stop here, they are yours to use with your own team or with anyone else.

The fee for the Discovery Sprint is on the pricing page. Travel and the prototype’s AI model usage are billed at cost.