The Basic ROI Formula
ROI = (Value created - Cost of implementation) / Cost of implementation, expressed as a percentage.
Value created = time saved x hourly cost, plus revenue gained, plus errors avoided.
Cost of implementation = platform subscription + LLM tokens + setup time + ongoing maintenance.
Simple enough. The hard part is quantifying the pieces honestly.
Quantifying Value Created
Time saved: multiply hours saved per week by the fully-loaded hourly cost of the person doing the work. Include benefits and overhead, not just salary.
Revenue gained: attribute new revenue to the workflow only where you can make a direct connection. Be conservative.
Errors avoided: estimate the cost of errors the old process was creating and multiply by error reduction rate.
Customer satisfaction: harder to quantify. Use NPS, CSAT, or retention as a proxy if it has moved.
Quantifying Costs
Platform subscription: straightforward. Monthly or annual platform fees.
LLM tokens: track token usage through the platform. At the end of each month, look at the actual spend.
Setup time: hours to design and build the workflow, multiplied by hourly cost.
Maintenance time: ongoing hours per week spent tweaking, debugging, and updating.
Opportunity cost of the person building the workflow: what would they otherwise be doing?
Example: Lead Response Workflow
Scenario: you have 200 inbound leads per month. Previously each took 10 minutes of a sales rep's time to process and respond.
Old cost: 200 x 10 / 60 = 33 hours per month. At $50/hour, that is $1,650.
AI workflow: drafts responses automatically. Rep spends 2 minutes reviewing each. New time: 200 x 2 / 60 = 6.6 hours. Cost: $330.
Savings: $1,320 per month.
Platform cost: $200 per month including tokens.
Net value: $1,120 per month. ROI: 560%.
Plus the upside of faster response times leading to better conversion. Measure that separately.
Common Mistakes
Forgetting LLM token costs. They can easily double platform costs at volume.
Ignoring setup and maintenance time. A workflow that costs 40 hours to build needs to pay that back before it starts creating net value.
Over-attributing revenue. Do not claim every new customer came from your AI workflow. Be honest.
Not tracking actual usage. The projected ROI only matters if the workflow actually runs at the expected volume.
Counting efficiency gains that do not translate to real savings. If you save 5 hours per week but do not reduce headcount or redirect to revenue-generating work, the savings are theoretical.
When ROI is Hard to Measure
Customer satisfaction: use NPS and retention as proxies.
Employee morale: use survey data or turnover metrics.
Compliance risk reduction: harder to quantify. Document incidents avoided where possible.
Speed: faster response times lead to more wins but the attribution is fuzzy. Compare conversion rates before and after.
Realistic ROI Expectations
Simple workflows (lead response, content drafting): 200 to 800% ROI in the first year.
Complex workflows (customer service, document processing): 100 to 400% ROI in the first year.
New use cases (experiments): often negative ROI in year 1, positive in year 2 if you scale.
Bad workflows: zero or negative ROI. Kill them fast.
Automate Operations and Get Discovered by AI
Agent Console HQ combines workflow automation with AI search visibility. Your AI agents do the work AND appear in ChatGPT, Claude, Perplexity, and Gemini when your customers search. Run your operations and your acquisition from one platform.
Explore Agent Console HQFrequently Asked Questions
Most workflows show positive ROI within 2 to 4 months. Complex enterprise deployments can take longer.
200% in year 1 is a reasonable target for most simple workflows. More complex work might be 100 to 150%.
Yes. Token costs are real money and often the biggest hidden cost. Track them monthly.
If it does not pay back setup and maintenance costs within 6 months, kill or rebuild it.
Estimate conservatively. Track for a few weeks before and after. Do not invent numbers, but do not let perfection prevent measurement.
Only if you do something with the freed time. Saving 10 hours per week is theoretical if nothing changes. Redirect it to revenue work or reduce hours.
Not in the core calculation. Track it separately as a qualitative factor. Hard to defend hard numbers you cannot justify.