Choosing How Your Reports Get Written
By Rob Brown, nSymbol Technology (~10 min read)
Every dynamic report you produce has to answer a question, usually several times over: who decides what this section says?
Take a client proposal. One section lists the services included in the package the customer chose. Another summarizes why this customer's situation calls for those services. The first is largely a matter of fact: if they chose the premium package, these items appear. The second needs to read like a person wrote it, tailored to the details of this particular customer.
These sections are different kinds of work, and they call for different tools. In Tag, we call the two approaches Logic-driven and AI-driven. Many reports use both. The skill is knowing which one belongs where.
Logic-Driven Content: Rules You Set Once
Logic-driven content follows fixed rules that you define with point-and-click tools. You decide the conditions once, and from then on the same data always produces the same output.
A few examples of what that looks like in practice:
Conditional sections. A "Special Terms" section appears only when the customer is outside the country. An inspection report shows the "Safety Concerns" section only when something was flagged.
Categories from numbers. A score below 70 is described one way, 70 to 85 another, and above 85 a third. Nobody types the description; the rule picks it.
Tables that grow with the data. An invoice has three line items or thirty, and the table simply gets as many rows as needed.
Lists built from the data. "Services included: setup, training, and ongoing support," with the commas and the final "and" handled correctly no matter how many items there are.
Names and pronouns. Once the person's name and gender are entered, every "he," "she," "they," "his," or "their" in the report follows automatically.
The strength of this approach is consistency you can defend. If someone asks why a section says what it says, you can point to the rule. Two people entering the same data get the same document. Nothing surprises you on page four.
The cost is up-front effort. Someone has to build the rules, and someone has to maintain them when your standards change. For content that rarely changes and has to be right every time, that is a good trade. For free-flowing narrative, it is not.
AI-Driven Content: A Draft Written From Your Data
AI-driven content works the other way around. Instead of writing rules, you collect the information and let AI draft the narrative from it.
In Tag, this happens through AI fields. An AI field looks like any other field on your form, except it comes with a few resources attached:
A prompt that tells the AI what to write and which of your other fields to draw on. You can read and edit it, so nothing is hidden.
A writing sample so the draft matches your voice and structure rather than sounding generic.
Optional attachments such as reference documents the AI should consider.
When you run the field, the AI reads the data you entered and produces a draft. Then it appears in the form like any other text, and you review it, adjust it, and move on.
There is one detail worth understanding because it removes a common worry: what happens to the formatting? AI models naturally write in a simple shorthand called Markdown, where a pound sign marks a heading and dashes mark list items. Tag converts that automatically into the real styles of your document. AI-written headings, bullet lists, and even tables pick up your fonts, spacing, and branding. In the finished report, the AI-written parts look exactly like everything around them.
The strength of this approach is speed and flexibility. Writing a tailored two-paragraph summary of someone's situation is exactly the kind of work AI does well, and exactly the kind that is tedious to build rules for.
The cost is variability. Two runs may produce slightly different wording, and the AI can be wrong. A person should always read the result before it goes out. (Tag also keeps a trace record of each AI request, including what it cost, so you can see exactly what happened.)
Hybrid: Using Each Where It Fits
Most real reports don't fit neatly into one camp, and they shouldn't have to. A hybrid report uses fixed logic for the parts that must be consistent and AI for the parts that benefit from fluent narrative.
Here is how that might look for a project status report:
Notice the pattern. The parts where an error would be embarrassing or costly are rule-based. The parts where the main challenge is putting facts into good prose are AI-drafted. In the finished document, a reader cannot tell which is which, and shouldn't need to.
What Both Approaches Depend On: Clear Data Fields
It is tempting to think the interesting decision is "logic or AI." In practice, a more foundational decision comes first: what information do you collect?
Both approaches read from the same place, your data fields. These are the pieces of information that make up your form: client name, interview date, scores, notes, selected options. The rules in a logic-driven section look at those fields. The prompt in an AI-driven section draws on those fields. If the fields are vague or missing, neither approach has anything solid to work with.
Two things make this less daunting than it sounds:
Your form is built for you. When you define the fields, Tag generates the entry form automatically: calendar pickers for dates, checkboxes for yes/no, dropdown lists for fixed choices, an "add another" button for anything that repeats. Change the fields and the form follows.
An AI field is just a field. Once the AI fills it in, it is treated like any other piece of data. The rest of your report doesn't care whether a value was typed, imported, or drafted.
A few habits keep your fields healthy over time: group related items together, keep the structure shallow rather than deeply nested, use "add multiple entries" for anything that repeats, and make fields optional unless the report truly can't exist without them. People often start a form with partial information, and unnecessary required fields only create friction.
Choosing: A Quick Decision Guide
When you are deciding how a particular section should be handled, these questions usually settle it.
Lean logic-driven when:
The content must be identical for identical data
You need to explain or defend exactly why it says what it says
The wording rarely changes from one report to the next
It involves calculations, thresholds, categories, or required statements
Lean AI-driven when:
The wording should be tailored to the individual
The section synthesizes many facts into narrative
Building rules for every variation would take longer than reviewing a draft
You are comfortable reading and editing the result before it goes out
Choose hybrid when: your report has both kinds of sections, which is most of the time.
Not sure? Ask Build with Tag
You don't have to work through this decision alone. Tag's chatbot assistant has a Build with Tag mode that helps you think it through together. The chatbot is available on Tag's dashboard, on all editors, and on the form viewer.
Describe the report you want to produce, and it can walk through the sections with you: which ones look like fixed rules, which ones suit an AI draft, and where a mix makes sense. It can then set up the report accordingly, whether that means building the fields and sections for you or starting from a ready-made layout when your subject area has one available.
If you already know what you want, you can skip the conversation and choose your approach directly when you start a new report. Either way, the choice isn't permanent. Because the sections are separate, you can start AI-driven and replace a section with fixed rules once you notice you are correcting the same thing every time, or the reverse.
Reuse: Where the Effort Pays Off
Whichever approach you pick for a section, the work doesn't have to be thrown away after one report. A well-built section, such as a standard header with client details, a score table, or a set of required statements, can be saved once and reused across many reports. Fix it in one place and every report that uses it gets the update.
For some subject areas, Tag also includes a catalog of ready-made field layouts and report content, so you can start from something close to what you need rather than a blank page. Custom catalogs can be made to fit your specific organization’s requirements.
Over time, this adds up. Your reports become a library of tested pieces, some rule-based, some AI-assisted, that reflect how your practice or business actually communicates.
Try It on One Section
The easiest way to see the difference is to pick a single section of a report you already produce. Ask yourself: if two different people filled this in with the same information, should the result be word-for-word identical, or should it read like it was written about that specific subject?
Your answer tells you where to start. Open Tag, start a new report, and either choose an approach or let Build with Tag help you decide. See Get started for a practical example.
Licensed under CC BY 4.0.
