×
Different clients have different goals. The AI applies the correct logic per account- not a one-size-fits-all approach.
AI-powered Google Ads search term analysis - built in-house for the Uplers team.
Across all account managers
Each with its own Sheet
Up from manual 0
Before this tool, reviewing search terms was done manually - hours per account, per week. Now it takes minutes. The AI analyses every term, assigns a decision, and writes results directly into the team member's Google Sheet.
Every account manager was spending hours reviewing search terms by hand. It was slow, inconsistent, and not all accounts could be covered every week.
2–4 hrs × 15–20 accounts
Across 50+ managers
Same logic across all accounts
No technical knowledge required. One-time setup, then three steps every time.
Double-click the installer, enter work email, done. The tool installs itself silently - no IT involvement, no admin rights needed. Starts automatically every time the laptop is switched on.
Each manager has their own Google Sheet per client. They open the sidebar, confirm the client goal settings, and click Run Analysis.
Two AI models work in sequence. The first classifies every term into one of 12 decisions. The second cross-checks uncertain results. Up to 100,000 terms per run.
Negatives, keywords to add, opportunities, and items for review - all written into separate tabs. Summary dashboard and history updated automatically.
The manager never leaves Google Sheets. No separate tool, no login, no export or import. Results appear where the work already happens.
Different clients have different goals. The AI applies the correct logic per account - not a one-size-fits-all approach.
Leads, Sales, Traffic, Awareness, App Installs, Local Visits - the AI understands what success looks like for each.
CPA target, ROAS target, CPC cap, CTR minimum - the AI evaluates every term against the client's actual number.
"Always negative competitor terms" or "exclude job-related searches" - written in plain English, applied automatically.
"Seasonal campaign Oct–Dec only" or "B2B only, ignore consumers" - the AI factors in everything the manager knows about the client.
Settings are saved per Sheet and set only once per client. A ROAS client and a CPA client with the same search terms will get completely different - and correct - recommendations automatically on every run.
AI API + a small cloud server fee. All infrastructure runs on each manager's own laptop - zero shared server cost.
Per 1,000 terms
~$0.08–0.12
Per 10,000 terms
~$0.80–1.20
Per 50,000 terms
~$4–6
Central server (Railway)
$5 / month
50 managers × 15–20 accounts ×
avg 5,000 terms × 1 run/week
~$300–600 / week
Monthly AI cost (all 50+ users, all accounts)
~$1,200–2,400 / month
Total monthly
~$1,205–2,405
$1,200 ÷ 50 managers ÷ 15 accounts
Less than the cost of a coffee per client account per month.
Conservative estimate at ₹200/hr
1,500–4,000 hrs saved/week × ₹200 = ₹3–8 lakh/week in recovered productivity. Tool costs ~₹1–2 lakh/month.
Runs on manager laptops
Built in-house
All accounts included
Next planned optimisation: direct Gemini API key removes the OpenRouter middleman commission (~15–20% cost reduction) + token optimisation reduces AI usage by a further 30–40%. Combined saving could bring the monthly bill down to ~$700–1,400.
Built iteratively - starting from a basic proof of concept, solving real scaling and reliability problems along the way.
AI analysis via Google Sheets sidebar. Two-pass AI (Gemini + GPT-4o-mini). 12 decision types. Results written back to Sheet tabs automatically.
Single shared office server crashed under 50 users. One shared API key hit rate limits - everyone blocked simultaneously. 70,000 terms taking 8–10 hours. Server going down stopped all 50 users at once.
Moved from shared server to each manager's own laptop - no more single point of failure. Per-user API keys — no shared rate limits. Central auth server on Railway for access control, daily limits, and budget tracking. Silent one-click installer. Auto-update system pushes new code to all machines silently.
Per-client goal settings: Campaign Objective, Goal Type (CPA/ROAS/CPC/CTR), Target Value, custom rules. AI applies different logic per account automatically.
Direct Gemini key, token optimisation, one-click push to Google Ads, scheduled automatic runs.
Four planned improvements - two reduce cost, two remove manual work entirely.
Currently routing through OpenRouter which adds a commission on every call. Switching to a direct Google Gemini key removes this entirely - estimated 15–20% cost reduction with zero change to output quality.
The data sent to the AI per batch can be trimmed without losing decision quality. Removing redundant fields and compressing the system prompt is estimated to reduce token usage by 30–40% - significantly cutting the monthly AI bill.
Today: manager reviews AI decisions, then manually applies them in Google Ads. Next: manager reviews, clicks Approve, changes apply automatically. One click replaces hours of implementation work per account.
Search campaigns: weekly automatic runs with no manual trigger — ready to build. PMax campaigns: waiting for Google to enable search term access via their API (not yet available). Once enabled, the entire workflow becomes fully automatic for all campaign types.
The admin has full visibility and control over the entire team - from one central panel. No access to individual machines needed.
Add or remove team members instantly. Turn any user on or off without touching their machine. New users are live within seconds of being added.
Set a maximum daily AI spend per user. Set a monthly budget cap for the entire team. If the budget is hit, the system pauses automatically - no surprise bills.
Cap how many analysis runs each user can do per day. Prevents accidental over-usage while giving the team full flexibility within the limit.
Every run is logged - who ran it, how many terms, how long it took, cost, status. Full audit trail visible in the admin Google Sheet in real time.
One toggle puts the entire system into maintenance mode - all users get a clear message, no jobs processed until the admin turns it back on.
Each user has their own AI API key stored securely in the admin sheet. No shared key - one user hitting a limit never affects another user's work.
All admin controls live in a single Google Sheet. No dashboard to log into, no code to change. Add a user, set a limit, check usage - all in the same place the team already works.
A production-grade in-house AI tool - live, used daily, and improving every sprint.
50+ users on v1.0.1
All 50+ users, all accounts
Less than a coffee
Full AI analysis · Per-client goal settings · 12 decision types · Auto-update to all machines · Central admin controls · Usage and spend tracking · Silent one-click installer · Per-user API keys · No shared server
Direct Gemini key (lower cost) · Token optimisation (lower cost) · Push to Google Ads (less manual work) · Scheduled runs for Search campaigns · Full automation when Google enables PMax
The end state: a manager opens Google Ads on Monday morning and the previous week's search terms have already been reviewed, decisions approved, and changes applied - automatically, for every client, every week.
Different clients have different goals. The AI applies the correct logic per account- not a one-size-fits-all approach.
Leads, Sales, Traffic, Awareness, App Installs, Local Visits - the AI understands what success looks like for each.
CPA target, ROAS target, CPC cap, CTR minimum - the AI evaluates every term against the client's actual numbers.
"Always negative competitor terms" or "exclude job-related searches" - written in plain English, applied automatically.
"Seasonal campaign Oct–Dec only" or "B2B only, ignore consumers" - the AI factors in everything the manager knows about the client.
Four planned improvements - two reduce cost further, two remove remaining manual steps entirely.
Removes the OpenRouter middleman, cutting AI costs by 15–20% with zero change to output quality.
Trimming data sent per batch reduces token usage by 30–40%, cutting the monthly AI bill further.
Manager reviews AI decisions, clicks Approve - changes apply automatically. One click replaces hours of implementation per account.
Weekly runs with no manual trigger needed. The end state: a manager opens Google Ads on Monday and last week's work is already done.
