HELM: what it does and what it won't do

Most AI tools answer confidently whether or not they know. HELM was built the other way around. This guide explains how it reasons, where you'll meet it, and the things it deliberately won't do.

What HELM is

HELM is an AI assistant specialized in Google Local Services Ads. It isn't a general chatbot that happens to know about ads. Its advice is grounded in a curated doctrine rather than generic internet knowledge:

Evidence first, advice second

HELM follows the same order every time.

1. Read the record. Leads, calls, spend, position and outcomes are reconciled before any recommendation.
2. Separate known from missing. If outcomes are unresolved, HELM says so before answering anything that depends on them.
3. Apply the doctrine. Advice is grounded in curated Local Services expertise and stays inside the 36 never-recommend rules.
4. Show the evidence. Every answer comes with the evidence behind it, so you can check the reasoning yourself.

Here's how that sounds in the CallRadius Signal demo. The demo account is a fictional roofing company, and HELM's answers there are scripted from synthetic evidence; no live account or model is queried.

This account is producing. The strongest observed value is in Roof Replacement, while repeat contacts are making Google's reported cost per lead look materially lower than the cost per distinct person.

It continues: "I also found five booked outcomes that are not yet matched to revenue. I will show that gap before answering any question that depends on return." Under the answer, it lists what it loaded: 30 daily performance rows, 78 distinct leads, 34 booked outcomes and 29 revenue matches.

Known vs missing, in practice

That second sentence matters more than the first. In the demo, 34 jobs were booked and 29 of them are matched to revenue. Any answer about return, such as whether a category pays back or whether to cut budget, depends on those five missing outcomes.

A tool that ignores the gap might quietly treat those five jobs as worth nothing, or assume an average. HELM names the gap first. In the demo's category economics view, HELM says roof replacement creates the most observed value and adds that two unresolved outcomes "can change the magnitude, not the current direction." That's the kind of statement HELM aims for: what the evidence supports, and how far.

Where HELM works

• CallRadius (managed). HELM performs expert review of managed Local Services accounts, so recommendations reach your LSA manager with their reasoning attached. Changes to your account follow the same approval rules as everything else in CallRadius.
• CallRadius Signal (self-serve). You ask HELM about your own account: the cost gap, your strongest category, whether to change budget.
• Reports. In Signal reports, HELM writes the executive read, with the evidence behind every number.

What HELM won't do

It won't act on a Signal account

Signal is read-only, enforced in our systems. It cannot change bids, budgets, campaigns or settings, and that includes anything HELM suggests. If HELM recommends a change, you decide whether to make it in your own Google account. More in Your data, read-only access and the archive.

It won't guarantee results

HELM explains evidence and recommends. It doesn't promise leads, calls, bookings, revenue or ranking, and nobody can. Results depend on Google.

It won't answer return questions as if the gaps weren't there

If outcomes are unmatched, HELM shows that before discussing return, budget or which category pays.

It won't cross its never-recommend rules

The register of 36 rules sets the boundary, whatever the question.

It won't know what isn't in the record

HELM reads the account evidence and the figures you enter, such as what a job was worth. If something isn't recorded, it can say the evidence is missing; it can't fill it in.

HELM on each Signal plan

Prices are per connected Local Services account, in US dollars, excluding applicable taxes. Every plan starts with a 14-day free trial with no card required, and enrollment is opening soon. For how to make five questions count, see Asking HELM good questions.

Common questions

Is HELM a general chatbot?

No. It's bound to a doctrine of Local Services expertise and to the evidence in your account record, and it shows that evidence with its answers.

Can HELM change my budget in Signal?

No. Signal is read-only and cannot change bids, budgets, campaigns or settings. HELM recommends; you decide.

Are the HELM answers in the demo real?

They're scripted from the demo's synthetic evidence for a fictional business. No live account or model is queried in the demo.

What happens when I use my five questions?

On the $29 plan, HELM is limited to five questions per month plus one deep review per complete data window. The $14 Unlimited HELM add-on, or Signal Pro, removes those limits.