Finding information in a spreadsheet does not mean it was read correctly or put in the right place. A review process that leads back to the source makes uncertain fields and extraction errors visible.
Treat validation as a separate task
Taking information from a PDF, scanned document or image into a spreadsheet can reduce a substantial part of repetitive work. Opening the output or seeing filled columns does not validate it. A date may have been misinterpreted, a document number may have lost its leading zero, or two rows may have been mixed together. The goal is to preserve the meaning of the source, rather than simply find empty cells.
Write down which documents will be processed and which fields will be extracted. Identify the file type, document version and page range. If a folder contains old and new copies, define how the correct source is chosen. Do not force different document types into one template: different layouts may need separate mappings or review steps. Define the expected output around the daily needs of the person who will use it.
Connect each row to its source
A reviewer should be able to identify the file and page behind each row. A job number, document identifier and page reference can establish that connection. Do not assume that a filename is unique; another copy may have the same name. Record the source version when needed. Finding the same information again matters more than knowing the extraction mechanism.
If a document contains several tables, identify the section or table location. Explain whether headings, footnotes and rows on continuation pages count as data. Preserve the relationship when one document produces several output records. A value that cannot be explained by returning to its source needs review, even if its cell is filled. Traceability can be included in the delivery scope for document-to-table work.
Map fields by their meaning
Create a short mapping that shows source headings next to destination columns. Does a date mean the document date, transaction date or expiry date? What unit does a quantity use? Similar headings do not necessarily have the same meaning. Define what happens when a field cannot be found: leave it empty, create an exception or stop processing.
Record any approved transformation rule. Changing a date format, adjusting number separators or trimming unnecessary spaces can preserve meaning. Guessing an uncertain value is a different operation. For example, silently replacing an unreadable character with a similar letter is unsuitable. Show uncertain information and genuinely missing information as different states; the user may need different decisions for them.
Check formats for numbers and identifiers
Decimal separators, thousands separators, currencies and measurement units require attention. The meaning of “1.250” depends on the format used. Automatically converting an identifier or document number to a number can remove leading zeros. A small-looking change can break matching with another file. Check the actual cell value as well as its display.
Totals are useful checks, but do not replace row-level review. A missing row and another row extracted twice can balance each other. If the source contains a total, identify which items it covers. Do not combine different units or currencies. Where rounding is acceptable, the process owner should define its limit and stage. Do not invent a new rule during review to conceal a discrepancy.
Make exceptions understandable
A review report should show the file, page, field, expected value and observed state together. Unreadable fields, missing pages, duplicate documents and incorrect templates are different causes. A generic error message is insufficient. Explain which record needs review, why, and what should happen next. After correction, show that the affected row was checked again.
Do not test only clean examples. Choose a small representative set covering rotated pages, different headings, long tables, missing fields and continuation pages from actual work. Establish expected results from human-verified sources. The automation acceptance test approach ties success to observable criteria.
Separate accepted, corrected and pending records at the end. The output user should know which values came directly from the source and which were transformed by an approved rule. A correction made during a trial is not automatically a rule for every later document. If a new rule is needed, describe its scope and test its outcome on different examples.
Caution: protect documents and access
Avoid unnecessary personal or commercial information in trials. Reduce or mask examples where possible, and limit access to source files and review reports to the necessary people. Before using an external tool, establish which data goes where. Also review source preservation, previous versions and how failed processing can be repeated. Prevent a review copy from being mistaken for the original record.
URIKEM starts with your actual documents and the spreadsheet you use today. Describe the fields you extract, the errors you catch manually and who approves the result. A small trial can assess the output, source links and exception records together. Tell us about your work; together we can define the scope of software for your company, trials with your own files and an approved delivery.
Frequently asked questions
Does a filled spreadsheet prove correct extraction?
No. Source references, field meaning, number formats and relationships between records also need checking.
Should unreadable information be completed automatically?
Without a previously approved rule, it should not be guessed. Show it as an exception that requires review.
Which documents should be used in a trial?
Use a small representative set covering clean examples and the different layouts and missing fields encountered in actual work.
Is there a job that keeps repeating at your company? Tell us about it and we will build the tool that does it.
Tell us what you needOther articles
How to build a notification preparation checklist?
Preparation, review, approval and submission are separate stages of a notification. A checklist suited to the work exposes missing fields and records who approved which version.
7 October 2026Read →How to reconcile records after a data transfer
An uploaded file does not prove that information reached the right record or field. Define the scope first, then compare source and destination records to expose omissions, duplicates and unexplained differences.
6 October 2026Read →How to keep a change log for an internal guide
An up-to-date guide should also make clear what changed, when and why. A concise change log helps staff find the correct version, owners track updates and teams preserve the context of earlier decisions.
6 October 2026Read →Why does the internal guide keep getting postponed?
Start with one task and identify the knowledge owner, review point and update process together.
5 October 2026Read →How do you prepare an automation acceptance test?
Test sample data, expected outcomes, interruptions and recovery in one acceptance checklist.
5 October 2026Read →Whose code should be for the software you print?
Working software is not the same as having another team maintain it. Discuss code, documentation, data and access delivery together at the beginning of the project.
4 October 2026Read →Five questions to ask when choosing a software developer
The lowest bid alone is not the correct selection criterion. Evaluate maintenance, delivery, data, initial release, and training questions through the same sample process.
4 October 2026Read →Data transfer between two systems: integration or manual?
Writing the same information to two applications works for a while. When work is repeated, the decision should be made not only by speed but also by the need for data accuracy, traceability and maintenance.
3 October 2026Read →We have ERP, is there still a need for a special vehicle?
While ERP manages master records, some work may remain in emails and spreadsheets. The specific tool decision is made by first examining the possibilities of the current system and the actual process gap.
3 October 2026Read →Monthly plan or one-off payment?
A lower first quote does not guarantee a lower cost over two years. Compare both options with the same scope, maintenance obligations and exit conditions.
2 October 2026Read →Will AI replace an employee?
An employee’s job contains many different kinds of task. Assess these separately to understand where AI and automation may help.
2 October 2026Read →Customers missing appointments: do reminders help?
A reminder brings a forgotten appointment back into view; it does not guarantee attendance. Collect accurate contact details, send a confirmation and follow up with a short reminder before the appointment. Measure completed appointments, rather than bookings alone.
1 October 2026Read →What happens if your website is hacked, and how can you prepare?
When a change looks suspicious, protect the website and its users, preserve records and establish the scope. Restoring a clean backup is not enough on its own: close the entry point, secure relevant access and test the service again. Regular maintenance supports this preparation.
1 October 2026Read →Who benefits from an appointment system?
An appointment system can help when customers need to reserve a time slot and arranging those slots by phone interrupts your work. First clarify service durations, shared resources and exceptions.
22 September 2026Read →What if AI makes things up?
A fluent answer is not verified information. A business tool needs more than well-written output: it needs a workflow that prevents actions based on incorrect or incomplete information.
22 September 2026Read →Can a spreadsheet with thousands of rows be entered into a system in bulk?
In most cases, yes. The hard part is not the entry itself but getting the data into the shape the system expects and verifying that every row was really saved. If the system offers no bulk route, the work goes through the screen, and that depends on the rules of that system.
21 September 2026Read →Why does preparing a quote take so long?
Writing the quote does not take long; the time goes into hunting for the price, copying and correcting an old quote, and rebuilding the layout every time. Once the price list sits in one place and the rules are written down, a quote comes out in minutes.
21 September 2026Read →Which jobs are suited to automation, and which are not?
The job best suited to automation is one that repeats often, whose rules can be written down and whose data is already digital. Work that needs judgement, is full of exceptions or has its data on paper needs another step first.
18 September 2026Read →What if the AI assistant on your site gives a wrong answer?
The risk is real, but manageable. An assistant's wrong answer usually comes from guessing about something it does not know. The solution is to draw a clear line from the start around what the assistant knows and does not know, and to have it hand over to a person when it does not know.
18 September 2026Read →Is ranking in Google the same as being mentioned by AI?
Partly. Both draw on the same foundation, but they do not reward the same thing: Google ranks a page, while AI reads several sources, writes an answer and decides for itself who to mention in it.
17 September 2026Read →Why doesn't ChatGPT mention my company?
The reason is rarely “the AI does not like us”. Usually it is a technical block, scattered company information, or the company not being mentioned anywhere else online. All three can be fixed.
17 September 2026Read →What should your site have so AI describes you correctly?
When AI describes a company, it uses the clearest sentences it finds on the site. The list below goes page by page through where those sentences should sit, how they should be written and the markup machines will read.
17 September 2026Read →What is GEO? Being visible in AI answers
GEO (Generative Engine Optimization) is the work of making sure that when ChatGPT, Gemini, Perplexity and Google's AI Overviews answer a question, they find your business, describe it correctly and cite it as a source. It does not replace SEO; it builds on top of it.
17 September 2026Read →Will an AI tool learn my company's data?
Short answer: not in a tool running through a commercial API. But that answer depends on how the setup was done, so it is worth knowing what goes where.
16 September 2026Read →When does a spreadsheet stop being enough?
Excel is not a bad tool; it handles most jobs and more. The problem is carrying on with it for months after the point where it stopped being the right fit.
16 September 2026Read →Off-the-shelf software or custom-built?
The right answer depends on how specific the work is to you. Work that everyone does the same way is cheaper off the shelf; work you do your own way is cheaper built for you.
16 September 2026Read →