How to use AI to support data entry to TRACE

We developed a skill to be used with the AI platforms you’re already integrating to take on that translation work. It handles the heavy lifting of reading and converting your documents, recording every assumption it makes, so your team can focus on reviewing the results and acting on what the data shows.

Download the skill here.

A skill is a reusable folder of instructions that teaches an AI platform how to perform a specialised, repeatable task to your exact standards.

This skill has been set up to speed up data entry. We recommend you always review the output to ensure the results accurately reflect the event.

What the skill does

The skill reads your documents, converting data into the TRACE template for you. It will prioritize the real data that sits in those files, then where there are data gaps, it will apply TRACE assumptions to the data.

For example, where there are drinks menus available, but no consumption figures, it might assume half a bottle of wine per person over dinner, or 1 coffee per person over coffee breaks in the schedule document.

The tool will record every time it makes an assumption, so your team can review the figures with confidence before anything is reported.

Common assumptions exist in the TRACE rule document here, so that you can better understand and interpret your data. We’ve also included tips on how to edit the skill using your own ‘in-house’ assumptions so that they get applied consistently across the whole team using TRACE.

Each time you apply the skill to your documents, you'll receive:

Output	What's included
A completed TRACE template	Data entered on the right tabs, with the template's formulas - you can now just copy and paste this directly into TRACE.
An assumptions log	This shows every figure that didn't come directly from a document, the rule applied and a confidence level (High, Medium or Low) - keeping your data auditable and repeatable.
A reasonableness and gaps check	Quick sense checks to highlight any potential gaps using the information you’d inputted about the event and the data you’ve provided.

We recommend you review the assumptions carefully before adding to TRACE, using the chat function in your AI platform to adjust numbers as needed to apply additional event specific assumptions.

Before you start

You'll need four things for your first run:

The skill. Download the TRACE skill here and follow the steps to get started with it.
The source documents for one event. PDFs, spreadsheets, Word files and screenshots all work.
Include the TRACE template in your uploads so that the skill always works with the most recent version or with your own custom team templates. (Access the templates here)

At the start of each event, the skill will ask a few questions. Having the answers ready will speed things up:

Question	Why it's asked
Event name, dates, city, country and number of attendees	To support with sense checking the outputs and highlighting gaps
Which TRACE tabs to complete, and whether each dinner or day needs its own tab in the TRACE template	This sets the layout of your workbook; think about how you might produce a report, what data is important to keep separate, what can be reported togethet
Any key assumptions that are specific to the event	So they're applied constantly across the data extraction and recorded for later reference

Review the results

Start with anything marked 'Low confidence' or flagged in the checks, as that's where most of the uncertainty sits. Are you happy with the assumptions made? Remember that in some instances, it might be more appropriate to report on the gap rather than on data with very low certainty.

If you'd like a choice changed, say so in the chat; for example "report breakfasts as cooked" or "assume 125 ml glasses for wine". Claude will apply the change everywhere it was used and log the decisions you’ve made. If you find yourself making the same correction every time, it might be worth considering editing the skill for your own ‘house rules’. More information on that below!

Customising the rules to your house assumptions

Our default rules give you a solid starting point, but many organisations have their own standards e.g. a travel policy on flight class, a drinks allowance per guest or a house glass size. Setting these once, centrally, keeps your data consistent across every event and everyone on your team. It also means no one has to correct the same thing in every chat.

These are the rules organisations most often customise:

Rule	Default	Why you might change it
T3 Flight class	Economy unless booked otherwise	Your travel policy puts long-haul in business
D2 Evening drinks	Dinner: 2 wine + 1 soft. A reception adds 1 cocktail + 1 beer	Your caterers' typical figures are different
D5 Glass and measure size	150 ml wine, 50 ml spirit	Local serving sizes, such as 125 ml or 25 ml
GT2 Hire vehicles	100 km per vehicle per day	Short city days, or long rural transfers
F5 Breakfast	Hot items = cooked; cold only = continental	You report all breakfasts the same way
F13 Serveware	Reusable at hotels and restaurants	Your venues often use single-use items
T8 Local travel	Rail, 10 km per person	You know your local attendees' usual commute

To set your own rules:

Open TRACE Event Data Entry - Default Rules, choose File > Make a copy, and rename it for your organisation.
Add a new column for "Your house rule" column next to the default. Keep the rule IDs as they are and add in your new rules.
Open SKILL.md in a text editor, find the rule by its ID in section 5 ("Default rules") and replace the wording with your house rule. Add "(house rule, v1.1)" at the end so it's clear in the logs. Please leave the review caveat in section 4 unchanged.
Test it on one past event and check the assumptions log shows your new wording.
Zip the folder and publish it as a new version.

If you'd rather not edit the file by hand, attach SKILL.md and your copy of the rules to a chat and ask it to update section 5 to match. We then suggest checking the result line by line against your rules document before publishing.

Using the skill in Claude

The skill works on all Claude plans: Free, Pro, Max, Team and Enterprise. It needs code execution switched on, as it uses this to read and write your TRACE workbook.

Install the skill

Switch on code execution. On Free, Pro or Max, go to Settings > Capabilities and turn on "Code execution and file creation" (in newer layouts, "Create and edit files"). On Team or Enterprise, an owner will need to check that code execution and Skills are both enabled under Organization settings > Plugins & skills > Policy.
Upload the skill. Go to Customize > Skills, click "+", choose "+ Create skill", then "Upload a skill", and select the zip file.
Turn it on in Customize > Skills. If a colleague has shared it with you, you'll find it under "Shared with you", greyed out until you enable it.

Run the skill

Start a new chat and attach the blank TRACE template and your event documents.
Tell Claude what you need. If it doesn't pick up the skill automatically, name it, for example: "Use the TRACE event data entry skill to enter this rooming list and these BEOs into the TRACE template."
Answer its questions. Each one comes with a suggested default, and "use the default" is a perfectly good answer.
You'll receive the completed workbook, a short summary that leads with the biggest risk or gap, and the reasonableness check. If you need the assumptions log as a separate document for your records, just ask.

For more detail on the settings, see Anthropic's guides: Use skills in Claude and Provision and manage skills for your organization.

Using the same approach in ChatGPT

ChatGPT can run the same instructions, either as a Skill or as a custom GPT. Which one is available to you depends on your plan. Either way, it needs file handling switched on to read your documents and create the workbook.

Route	Who can use it	How to set it up
Skills in ChatGPT	Eligible Business, Enterprise, Healthcare and Edu users	Go to Skills > Create > Upload from your computer to upload the skill folder, or Create with editor and paste in SKILL.md. To share, go to Plugins, open the Skills tab, click ••• and choose Share. Admins manage sharing in Permissions & roles.
Custom GPT	Workspaces that still allow GPT creation (personal plans no longer do)	Go to Explore > GPTs > Create > Configuration view. Paste SKILL.md into Instructions, upload your rules document as a Knowledge file and turn on "Code Interpreter & Data Analysis". To share, go to My GPTs > Edit GPT > Share.

House rules work slightly differently in each. With Skills, update section 5 of SKILL.md as described above and upload the new version. With a custom GPT, paste sections 1–4 and 6–8 of SKILL.md into Instructions, upload your rules document as a Knowledge file, and add this line to the Instructions: "Where the uploaded rules file gives a house rule, it overrides the default rule with the same ID." When a rule changes, replace the Knowledge file and click Update.

Before your team relies on it, run one past event through it and compare the results with a Claude run or with entries you've made yourself. Check that the downloaded workbook keeps its TRACE dropdowns and formulas, and that the assumptions log still opens with the review caveat.

For more detail, see OpenAI's guides: Skills in ChatGPT and Creating and editing GPTs.

Using the same approach in Gemini

In Gemini, the closest equivalent is a Gem: a set of saved instructions with up to 10 knowledge files. Gems work particularly well for house rules. If you attach the rules document directly from Google Drive, the Gem always reads the latest version, so when the owner updates a rule, everyone has it on their next run.

To set one up:

Create a new Gem and paste sections 1–4 and 6–8 of SKILL.md into its instructions. Add one line: "Use the attached rules document for all default and house rules. A house rule overrides the default with the same ID."
Attach your copy of the rules document from Google Drive. Link to it rather than uploading a downloaded copy, otherwise later updates won't come through.
Save the Gem and test it on one past event.

To share it, open the Gems list in the sidebar, click Share next to the Gem, and add people or groups. Viewers can use it and see its instructions and files; Editors can also change them. Your Workspace admin controls whether Gem sharing is switched on.

As updates to the rules document reach everyone straight away, we recommend keeping edit access with the skill owner only.

During your test run, check whether the Gem can return a completed .xlsx workbook, as Google's documentation doesn't confirm this. If it can't, ask for the entries as one table per TRACE tab and copy them into the template yourself, keeping the dropdown values exactly as written. Make sure the assumptions log, confidence levels and review caveat still come through.

For more detail, see Google's guides: Share a Gem and Gems with knowledge files.

Download the skill | View the rules

Need more support? Our team is here to help. Get in touch via support@traceyour.events