Example Audit · Specimen document
Where AI fits in a wholesale business.
Four workflows mapped, sixty three hours a week accounted for, five opportunities weighed, and one recommendation on what to build first.
Example Audit
What an Audit actually finds.
The document a client receives, in full and on this page. Four workflows mapped to the hour, five AI and automation opportunities weighed against what they cost to build, and one recommendation on what to start with.
This is a specimen. The structure, method and level of detail match what a client receives. The company, its systems and its figures are composites, so the findings could be shown whole rather than redacted. Aivonto.com does not publish client work.
The Audit, page by page
Aivonto.com · Example Audit
Contents
Read this first
This is a specimen. Its structure, method and level of detail match the document a client receives. The company, its systems and its figures are composites, so the findings could be shown in full rather than redacted.
What an Audit hands you at the end.
Not a list of tools, and not a slide about what AI could theoretically do. A record of how the work runs today, what each part of it costs in hours, and a recommendation you can act on or ignore.
- 01What we foundThe short version, in numbers
- 02How this Audit was runWho we spoke to, what we measured, how the estimates were built
- 03How the work flows todayFour workflows, step by step, with the hours each one takes
- 04Where AI helps, and where it does notIncluding the parts we would leave alone
- 05The opportunities, weighedFive of them, ranked on value, effort and confidence
- 06What to build firstThe recommendation, in detail, including what it will not do
- 07Assumptions and risksWhat has to be true for the numbers to hold
- 08Where this leaves youYour options from here, including stopping
Example Audit
02
01 · What we found
Aivonto.com
01
Most of the lost time sits in one place.
The measured week
Figures are team totals per week, averaged over three weeks in September and cross checked with the people doing the work.
Across the four workflows we mapped, your team spends 63.5 hours a week on work that is largely mechanical: reading documents, retyping what they say into another system, and looking things up to answer a question that has already been answered before.
Roughly 30 of those hours a week can be taken out with technology that exists today and integrates with the systems you already run. Not all at once, and not from every workflow. The bulk of it sits in one place: order intake.
The other finding matters just as much. Two of the things on your own wish list, the customer service inbox and stock forecasting, are not worth building yet. Section 04 explains why, and what would have to change first.
63.5 h
Spent each week on the four workflows we mapped
27 to 37 h
Realistically removable each week, once built and settled
~1,400 h
Per year, at 46 working weeks. Worth about €53,000 at your loaded rate
What those hours are, and are not
Recovered hours are not a cost saving on their own. Nobody leaves, and the payroll does not change. What you get back is capacity: two inside sales people who can handle a busier season without a third hire, and quotes that go out the same day instead of two days later. The euro figures in this document are what that time is worth if you put it back to work. Where we quote a payback period, we assume you only manage to use half of it.
The recommendation
In full in section 06.
Build the order intake assistant first. It accounts for over half of the recoverable time, it touches one system rather than four, and it can run alongside the current process while it proves itself.
01 · What we found
03
02 · How this Audit was run
Aivonto.com
02
Where the numbers in this document come from.
Why this section exists
An estimate you cannot trace is a guess with a decimal point. Everything in section 05 can be traced back to one of the four sources listed here.
The Audit ran over two weeks. Most of what is useful in it came from watching the work happen and asking the people who do it, not from a questionnaire.
- Two half days on site. Six people interviewed by role: two inside sales, one warehouse lead, one finance administrator, the commercial manager, and the owner.
- Three weeks of volume data. Message counts from the shared order mailbox, order line counts from the ERP, and supplier invoice counts from the finance inbox.
- Timed samples. Forty orders, fifteen quotations and twenty five supplier invoices, timed end to end by the people processing them.
- A systems review. What your ERP, mailbox and document storage can and cannot be connected to, confirmed against their documentation rather than assumed.
How to read the estimates
| Term | What it means here | Precision |
|---|---|---|
| Hours per week | Team total across everyone who touches the step, not per person | Within about 15% |
| Removable | The share of a step a system can take over, with a person still approving the result | A range, not a figure |
| Value | Removable hours at €38, your loaded hourly cost including employer charges | Indicative |
| Effort | Build and integration work, in weeks, before anything goes live | Indicative |
| Confidence | How sure we are, given what we could verify in two weeks | High, medium or low |
What we could not verify. Your ERP supplier had not confirmed API access at the time of writing, so section 07 treats it as an open question rather than a given. The volume sample also falls in September, which your team describes as an average month. If it is not, the annual figures move with it.
02 · How this Audit was run
04
03 · How the work flows today
Aivonto.com
03
Four workflows, drawn to the hour.
How to read a ledger
Every step is a bar, scaled to the hours it takes each week. The copper part is what a system can take over. The dark part stays with a person, either because judgement is needed or because the input is not machine readable.
Each workflow below is the process as your team described it, not as the procedure says it should run. Where the two differed, we drew what actually happens. Bars are on one shared scale, so a long bar in one workflow is directly comparable to a long bar in another.
A system can take this over
Stays with a person
Order intake
3 people · 26.5 h per week · ~240 orders
Sorting and assigning incoming order mailInside sales
3.5 h
Reading order documents, matching article numbersInside sales
8.5 h
Keying order lines into the ERPInside sales · admin
7.0 h
Checking stock and confirming delivery datesInside sales · warehouse
4.5 h
Confirming orders and chasing unclear onesInside sales
3.0 h
Removable: 14 to 18 hours a week. Two thirds of this workflow is reading a document and retyping it. About one order in eight arrives in a format nobody has seen before, and those stay manual either way.
Quotations
2 people · 17.0 h per week · ~35 quotes
Finding comparable past quotesInside sales
3.0 h
Pricing and margin checksCommercial manager
5.0 h
Writing and formatting the quoteInside sales
6.0 h
Following up on open quotesInside sales
3.0 h
Removable: 6 to 9 hours a week. Pricing stays human. Assembling the document around the price does not.
03 · How the work flows today
05
03 · How the work flows today
Aivonto.com
Supplier invoice handling
1 person · 9.0 h per week · ~130 invoices
Collecting invoices from mail and portalsFinance admin
1.5 h
Matching invoice lines to purchase ordersFinance admin
5.0 h
Entering and routing for approvalFinance admin
2.5 h
Removable: 5 to 7 hours a week. The smallest workflow, but the highest proportion removable, because the matching rule is already written down and consistently applied.
Customer service inbox
4 people, shared · 11.0 h per week · ~310 messages
Reading and triaging messagesShared
4.5 h
Looking up order and delivery statusInside sales
4.0 h
Writing repliesShared
2.5 h
Removable: 2 to 3 hours a week. Low, and the reason is in section 04. The questions look repetitive but the answers are not.
The week, totalled
| Workflow | Hours | Removable | Share |
|---|---|---|---|
| Order intake | 26.5 h | 14 to 18 h | About 60% |
| Quotations | 17.0 h | 6 to 9 h | About 45% |
| Supplier invoice handling | 9.0 h | 5 to 7 h | About 70% |
| Customer service inbox | 11.0 h | 2 to 3 h | About 25% |
| Total | 63.5 h | 27 to 37 h | About 50% |
03 · How the work flows today
06
04 · Where AI helps, and where it does not
Aivonto.com
04
The honest half of the answer.
Why this matters
The quickest way to lose money on AI is to build something for a process that was never the problem. Two of the four candidates here fall into that category.
The pattern across your four workflows is consistent. Where a document arrives in a predictable shape and has to be turned into structured data, automation works well and is cheap to prove. Where the work is deciding something, it does not, and dressing it up as an assistant just adds a screen for someone to check.
Where it genuinely helps
- Reading order documents. The formats vary but the content does not. Extraction is reliable and the failure mode is visible, not silent.
- Matching text to article numbers. A customer writing "16mm copper elbow" instead of your code is exactly the gap this technology closes.
- Drafting structured records. Order lines, quote documents, invoice bookings: the system proposes, a person approves.
- Matching invoice lines to purchase orders. A written rule, applied consistently, at volume.
- Assembling a quote around a decided price. The writing, not the pricing.
Where we would leave it alone
- The customer service inbox. We read 120 messages. Most need a live fact from the ERP plus a judgement call about a specific customer. An assistant would draft a reply someone has to verify anyway.
- Stock forecasting. Your historical data has two years of gaps from the system migration. A forecast built on it would be confident and wrong.
- Pricing and margin decisions. These are commercial calls with customer history behind them. Automating them transfers risk without removing work.
- Phone orders. About 20% of intake. Recording and transcribing them is possible, but the error cost is high and the volume does not justify it yet.
- The stock and delivery date check. It is a conversation with the warehouse, not a lookup.
What would change our mind
The service inbox becomes worth building once order status can be read live from the ERP, because then a draft reply contains a fact rather than a placeholder. That is the same integration the order intake assistant needs, so it may arrive as a side effect. Stock forecasting becomes worth revisiting after roughly eighteen months of clean post migration data, which puts it in 2028 at the earliest.
04 · Where AI helps, and where it does not
07
05 · The opportunities, weighed
Aivonto.com
05
Five candidates, and the order to take them in.
Value
Removable hours a year at €38 loaded cost.
Effort Build and integration weeks before anything runs live.
Effort Build and integration weeks before anything runs live.
Everything we found is plotted below on the only two axes that decide the order of work: what it is worth in a year, and what it takes to build. Confidence is carried in the table, not the plot, because an uncertain estimate should not look precise.
| # | Opportunity | Value a year | Effort | Confidence |
|---|---|---|---|---|
| 1 | Order intake assistant | €26,000 to €31,000 | 5 to 7 weeks | High |
| 2 | Quote drafting | €12,000 to €16,000 | 4 to 5 weeks | Medium |
| 3 | Supplier invoice matching | €9,000 to €12,000 | 2 to 3 weeks | High |
| 4 | Service inbox assistant | €4,000 to €6,000 | 4 weeks | Low |
| 5 | Stock forecasting | Unknown | 10 weeks + | Low |
05 · The opportunities, weighed
08
05 · The opportunities, weighed
Aivonto.com
The three worth doing
01
Order intake assistant
Build firstOrder documents arriving in the shared mailbox are read, the lines are matched against your article master, and a draft order is prepared in the ERP for inside sales to approve. Anything the match is unsure about goes to a review queue instead of into the system.
Value a year
€26,000 to €31,000
Build effort
5 to 7 weeks
Confidence
High · volumes measured
02
Quote drafting
Worth doing afterA request comes in, comparable past quotes are surfaced automatically, and the document is assembled once the commercial manager has decided the price. The decision stays where it is. The forty minutes of assembly around it does not.
Value a year
€12,000 to €16,000
Build effort
4 to 5 weeks
Confidence
Medium · 15 quotes timed
03
Supplier invoice matching
Worth doing afterInvoices are collected from mail and supplier portals, matched line by line against the purchase order, and booked for approval. Only mismatches reach a person. The smallest of the three builds, and the one with the cleanest rule behind it.
Value a year
€9,000 to €12,000
Build effort
2 to 3 weeks
Confidence
High · rule already written
Held back, and why
| 04 · Service inbox | The questions repeat, the answers do not. Revisit once the ERP integration from opportunity 01 makes live order status available to a draft reply. |
| 05 · Stock forecasting | Two years of gaps in the sales history from the system migration. Revisit after roughly eighteen months of clean data, so 2028 at the earliest. |
05 · The opportunities, weighed
09
06 · What to build first
Aivonto.com
06
The order intake assistant.
Why this one
It holds over half the recoverable time, it touches one system rather than four, and it can run beside the current process until it has earned trust.
Orders arrive as PDFs and email text in the shared mailbox, and three people spend most of their week turning them into ERP lines. That is the single largest block of mechanical work in the business, and it is the kind of work this technology is genuinely good at.
The week this workflow costs, before and after
TodayThree people
26.5 h
Once liveApproval and exceptions
~10 h
How it would work
- An order lands in the shared mailbox and is picked up automatically, PDF or plain text.
- The lines are read out of the document: article description, quantity, reference, requested date.
- Each line is matched against your article master and given a match score.
- A draft order appears in a review screen with the source document beside it, matched lines in green and uncertain ones flagged.
- Inside sales approves or corrects, and only then is the order written to the ERP. Every correction is kept, so the matching improves on the terms your customers actually use.
What it deliberately will not do
- Write to the ERP unattended. Not in the first phase, and not until the match rate has been stable for a quarter.
- Handle phone orders. Out of scope. They stay exactly as they are.
- Check stock or promise a date. That conversation with the warehouse stays with a person.
- Invent an article. No confident match means a flag, never a guess.
- Replace anyone. The design assumes the same team, handling more.
5 to 7 wks
From kickoff to running alongside the current process
€7k to €10.5k
Indicative build, fixed in a proposal before anything starts
~7 months
To pay for itself, assuming only half the recovered time is used
06 · What to build first
10
07 · Assumptions and risks
Aivonto.com
07
What has to be true for these numbers to hold.
Before a proposal
The two items marked as open below are checked first. If either fails, the recommendation changes and we will say so rather than work around it.
Every figure in this document rests on something we either measured, were told, or could not yet confirm. Here they are, separated.
Assumptions
| What we assumed | Basis | Status |
|---|---|---|
| Order volume of about 240 a week | Measured, 3 weeks | Confirmed |
| September is a representative month | Told by the team | Accepted |
| Loaded hourly cost of €38 | Given by finance | Confirmed |
| 46 productive weeks a year | Holiday schedule | Confirmed |
| Your ERP allows order creation through an API or import file | Not yet answered by the supplier | Open |
| The article master is complete enough to match against | Sampled, 12% gap found | Open |
Risks, and what we would do about them
| Risk | If it happens | Likelihood |
|---|---|---|
| The ERP has no usable interface | Fall back to a generated import file, which your team uploads. Adds roughly a week and leaves about 3 hours a week on the table. | Medium |
| Article matching stays below 85% | The review step takes longer than modelled and the saving lands nearer 11 hours than 16. Still worth building, on a longer payback. | Medium |
| The team keeps working around it | The most common reason automation fails. Handled by involving the two people who do the work in the review screen design, not by training them afterwards. | Low to medium |
| A large customer changes document format | Extraction quality drops for that customer until the template is retrained. A day of work, covered under maintenance. | Low |
On the build figure. The €7,000 to €10,500 range is indicative, based on the scope described in section 06 and on the ERP question resolving in the ordinary way. An implementation is quoted at a fixed price in a separate proposal, after those two open items are closed. Nothing in this Audit commits you to it.
07 · Assumptions and risks
11
08 · Where this leaves you
Aivonto.com
08
Three ways forward, and one of them is nothing.
The Audit is finished and settled on its own. What happens next is entirely your decision, and stopping here is a normal outcome we have no problem with.
Option A
Take it in house
Everything in this document is yours: the workflow maps, the figures, the reasoning. Your own IT partner can build from it, or you can shelve it. No part of it is locked to us.
Option B
Build the recommendation
We close the two open items in section 07, then quote the order intake assistant at a fixed price. If either item fails, we tell you before you spend anything on a build.
Option C
Start with the small one
Supplier invoice matching is two to three weeks and the cleanest rule of the three. A smaller way to find out whether working like this suits you.
Consulting firm helping businesses through automation, implementation & strategy.
Questions about this Audit
Aivonto.com@gmail.com
aivonto.com
About this specimen
A composite example, matching the structure and detail of a real Audit. Figures illustrate the method rather than describe any one client.
Example Audit · Aivonto.com
12
Left and right arrow keys turn the page too.
This is what €395 buys, for your business.
An Audit of your own workflows, at a fixed fee, booked on its own. It starts with a free call to work out whether there is a case worth pursuing at all.