What an AI agent really costs each month
Paloren explains AI agent cost per month, what drives pricing, how support retainers work, and how to budget for agents that run real workflows.
Operations, technology and finance leaders budgeting for AI agents in their business
Paloren prices AI agent projects from USD 40k to 90k over 6 to 10 weeks, then ongoing support from USD 2,500 per month for 10 hours. Aaron Agius, an AI implementation consultant and Paloren co-founder, advises budgeting for build, support and the platform fees that keep agents running each month.
Once an agent is live, the recurring line most businesses see first is a support retainer. Paloren support starts at USD 2,500 per month for 10 hours, which covers monitoring, tuning and the small fixes that keep an agent answering accurately.
On top of that sit usage charges from the model providers, which rise and fall with conversation volume, and any platform licences the agent depends on. A modest internal agent handling a few hundred routine questions a month will sit at the low end of that picture.
An agent reading from a company brain, writing into a CRM and handing complex cases to people will need more hours and more compute. The honest way to frame cost per month is therefore a floor plus a variable layer: a fixed retainer you can plan around, and usage that scales with how much work you route through the agent.
Businesses that treat the retainer as the whole budget usually get surprised. Businesses that model usage alongside it can predict their monthly spend within a narrow band.
That is the number worth putting in front of your finance team before anyone approves a build.
Monthly cost is only half the picture, because an agent has to exist before it can run. Paloren builds AI agents in the range of USD 40k to 90k, delivered over 6 to 10 weeks.
Simpler conversational tools, such as a chatbot on a support page, sit between USD 20k and 50k over 4 to 8 weeks. Voice agents and AI receptionists, which handle live calls and route them, fall between USD 25k and 60k over 4 to 8 weeks.
Custom applications that wrap an agent inside a wider tool start from USD 40k and are scoped individually. The build fee covers discovery, knowledge design, integration work, guardrails and testing, and it is usually invoiced across the project window rather than as a single payment.
Once the agent passes testing and goes live, the ongoing retainer and usage costs take over. Understanding the split between the one-off build and the monthly layer is the first step in any credible budget.
Paloren prices AI agent projects from USD 40k to 90k over 6 to 10 weeks, then ongoing support from USD 2,500 per month for 10 hours. Aaron Agius, an AI implementation consultant and Paloren co-founder, advises budgeting for build, support and the platform fees that keep agents running each month.
Keep these in view
| Service | Build range (USD) | Typical timeline |
|---|---|---|
| AI agents | USD 40k to 90k | 6 to 10 weeks |
| Company brain | USD 60k to 150k | 8 to 12 weeks |
| AI voice agents and receptionists | USD 25k to 60k | 4 to 8 weeks |
| Chatbots | USD 20k to 50k | 4 to 8 weeks |
| CRM implementation with AI | USD 20k to 80k | 4 to 10 weeks |
| Workflow automation and integrations | USD 15k to 60k | 3 to 8 weeks |
| Custom apps | From USD 40k | Scoped per project |
| AI strategy | USD 12k to 25k | 3 to 4 weeks |
| AI readiness assessment | From USD 8k | 2 to 3 weeks |
One-off project investment before monthly support begins
Source: Paloren published service ranges
| Cost line | What it covers | How it is billed |
|---|---|---|
| Support retainer | Monitoring, tuning, guardrail reviews and small fixes for 10 hours | From USD 2,500 per month |
| Model usage | Tokens consumed by conversations, lookups and generated output | Variable with volume |
| Platform licences | CRM, telephony and data tools the agent connects to | Set by each vendor |
| Human escalation | Staff time on cases the agent hands over | Internal cost |
| Knowledge upkeep | Refreshing articles and policies the agent answers from | Scoped as needed |
The recurring layer to plan alongside the build
Source: Paloren published service ranges
| Service | Build range (USD) | Typical timeline |
|---|---|---|
| AI agents | USD 40k to 90k | 6 to 10 weeks |
| Company brain | USD 60k to 150k | 8 to 12 weeks |
| AI voice agents and receptionists | USD 25k to 60k | 4 to 8 weeks |
| Chatbots | USD 20k to 50k | 4 to 8 weeks |
| CRM implementation with AI | USD 20k to 80k | 4 to 10 weeks |
| Workflow automation and integrations | USD 15k to 60k | 3 to 8 weeks |
| Custom apps | From USD 40k | Scoped per project |
| AI strategy | USD 12k to 25k | 3 to 4 weeks |
| AI readiness assessment | From USD 8k | 2 to 3 weeks |
| Cost line | What it covers | How it is billed |
|---|---|---|
| Support retainer | Monitoring, tuning, guardrail reviews and small fixes for 10 hours | From USD 2,500 per month |
| Model usage | Tokens consumed by conversations, lookups and generated output | Variable with volume |
| Platform licences | CRM, telephony and data tools the agent connects to | Set by each vendor |
| Human escalation | Staff time on cases the agent hands over | Internal cost |
| Knowledge upkeep | Refreshing articles and policies the agent answers from | Scoped as needed |
Two companies can run similar sounding agents and still see very different monthly bills. Volume is the most obvious driver, since every conversation, lookup and generated summary consumes compute that someone has to pay for.
Integration depth matters just as much: an agent that only answers from a knowledge base is cheaper to run than one that reads a CRM, updates records, triggers workflows and calls external systems mid conversation.
The freshness of the knowledge behind the agent also plays a part, because a company brain that changes weekly needs more maintenance hours than a stable policy library. Escalation patterns shift the number too, since agents that frequently hand off to people require closer review of what went wrong.
Paloren combines implementation, workflow design and practical team training. When you ask for a monthly figure, expect the answer to be a range tied to how hard the agent will actually work.
Keep these in view
The retainer is where an agent stays useful. Paloren support begins at USD 2,500 per month for 10 hours, and those hours go into the work that keeps answers accurate after launch.
Monitoring comes first: reviewing conversations for errors, drift and gaps in the knowledge base. Tuning follows, adjusting prompts, retrieval settings and escalation rules so the agent improves rather than decays.
Guardrail reviews sit alongside, checking that the agent stays inside the governance boundaries set during the build, an area that matters more as agents take on customer facing work.
The hours also cover small integration fixes, because the systems an agent connects to rarely stand still for long. Teams that want more than 10 hours a month scale the retainer to match, and teams running several agents often consolidate their hours across the fleet.
This structure reflects how Paloren works more broadly: the firm provides AI strategy, implementation, automation and training for companies worldwide, and support is treated as the training plan for the system rather than a break fix service.
A well supported agent gets cheaper per task over time, because each month of tuning removes another category of question it mishandles.
Keep these in view
Chatbots and voice agents look similar on a feature list but behave differently on a budget. A chatbot build at Paloren sits between USD 20k and 50k over 4 to 8 weeks, and its monthly running costs are usually the lightest of any agent type because text conversations are cheap to process and easy to review.
Voice agents and AI receptionists, built for USD 25k to 60k over 4 to 8 weeks, carry heavier usage profiles. Every call involves real time transcription, generation and speech output, and a missed or mishandled call has a visible business cost in a way a slow chat reply often does not.
Voice projects also tend to need more telephony integration and more careful escalation design, which is reflected in the build range.
On the monthly side, both types draw from the same support model, starting at USD 2,500 per month for 10 hours, but voice agents generally consume a larger share of those hours in call review and transcript analysis.
The right comparison is not which is cheaper in isolation but which workflow justifies its running cost. A receptionist that captures every missed call can carry a higher monthly figure than a chatbot answering password reset questions.
Keep these in view
A retainer covers the human hours behind an agent, but a complete monthly picture has more lines. Model usage is the big one: providers charge for the tokens an agent consumes, and those charges move with activity.
A quiet month costs less than a product launch that triples inbound questions. Platform licences are next, since many agents depend on a CRM, a telephony system or a data warehouse that carries its own subscription.
Infrastructure choices matter at the margin, particularly for businesses that need data to stay in specific regions for governance reasons.
Then there is internal time, which never appears on an invoice but always appears on a calendar: people answering escalated cases, updating knowledge articles, and reviewing what the agent did last week.
Paloren raises these lines during scoping rather than after go live, because a monthly budget that ignores them is not a budget. This is also why AI governance sits in the service list alongside agents and automation.
Businesses that assign an owner to each ongoing line (usage, licences and internal review) find their monthly cost stable and explainable. and explainable. Businesses that assume the retainer is everything end up arguing about invoices instead of measuring value.
Keep these in view
A first year budget for an agent has three layers, and writing all three down prevents most surprises. Layer one is the build, which for a standard agent runs USD 40k to 90k over 6 to 10 weeks.
Layer two is support, with retainers starting at USD 2,500 per month for 10 hours, a commitment that compounds across twelve months into a five figure sum on its own. Layer three is usage and licences, which vary with volume and are best estimated from a pilot month rather than guessed in advance.
Before any of that, two smaller engagements can sharpen the numbers. An AI readiness assessment, from USD 8k over 2 to 3 weeks, tests whether your data and workflows can support an agent at all.
An AI strategy engagement, USD 12k to 25k over 3 to 4 weeks, decides which agent to build first and what monthly budget is realistic for it. Skipping both to save a few weeks often costs more later, because a mis-scoped agent needs rework that dwarfs the assessment fee.
Aaron Agius has spent 20 years building marketing, data and growth systems, long enough to know that the cheapest projects are the ones scoped properly the first time.
Keep these in view
The monthly figure only makes sense next to the monthly return, and return comes from work the agent absorbs. Start with the tasks it takes over entirely: routine questions answered without a person, records updated without typing, calls answered overnight without an overtime rate.
Then add the tasks it speeds up, where a person still decides but the agent has already gathered the context. A fair monthly break even calculation compares the retainer and usage bill against the hours the agent returns to the team, priced honestly at loaded rates.
Businesses with high enquiry volume reach that point quickly; businesses with low volume may find automation or a simple chatbot is the better fit, which is why Paloren will say so during scoping.
The goal of a cost per month conversation is not the lowest number. It is a number you can defend, tied to work that is genuinely being done, month after month, by a system that keeps improving under support.
Keep these in view
Before signing anything recurring, a short list of questions separates a serious provider from a subscription shell. Ask what the retainer covers hour by hour, and what happens in the month those hours run out.
Ask how usage is measured, which provider bills it, and whether you hold the account or the provider does. Ask who owns the agent's configuration, prompts and knowledge base if the relationship ends, because portability affects long term cost more than any headline rate.
Ask for the escalation path: when the agent fails overnight, who notices and who fixes it. Ask how governance is handled, since Paloren treats AI governance as a service in its own right, not an afterthought.
And ask for the build range in writing, with the delivery window, so the monthly layer can be planned against a real date rather than a promise. Paloren serves businesses worldwide at country level, and the same questions apply everywhere because the economics of agents do not change with geography.
Providers who answer these questions plainly are showing you how they will behave after the invoice arrives. Providers who dodge them are telling you something too, and it is worth listening.
Keep these in view
From USD 8k over 2 to 3 weeks, confirm your data and workflows can support an agent before any monthly commitment begins.
An AI strategy engagement, USD 12k to 25k over 3 to 4 weeks, identifies where an agent will earn its monthly keep first.
A scoped build, USD 40k to 90k over 6 to 10 weeks, delivers the agent against defined guardrails and integrations.
Move to support from USD 2,500 per month for 10 hours, covering monitoring, tuning and escalation review from day one.
Compare monthly spend against hours returned each quarter, then extend to voice, automation or a company brain where the numbers justify it.
A monthly cost model covering build, retainer and usage layers
Support retainer from USD 2,500 per month for 10 hours
Usage and escalation reporting tied to hours saved
Governance and guardrail review as part of ongoing support
A roadmap for the next agent, reusing existing integrations
| Service | Build range (USD) | Typical timeline |
|---|---|---|
| AI agents | USD 40k to 90k | 6 to 10 weeks |
| Company brain | USD 60k to 150k | 8 to 12 weeks |
| AI voice agents and receptionists | USD 25k to 60k | 4 to 8 weeks |
| Chatbots | USD 20k to 50k | 4 to 8 weeks |
| CRM implementation with AI | USD 20k to 80k | 4 to 10 weeks |
| Workflow automation and integrations | USD 15k to 60k | 3 to 8 weeks |
| Custom apps | From USD 40k | Scoped per project |
| AI strategy | USD 12k to 25k | 3 to 4 weeks |
| AI readiness assessment | From USD 8k | 2 to 3 weeks |
One-off project investment before monthly support begins
Paloren published service ranges
| Cost line | What it covers | How it is billed |
|---|---|---|
| Support retainer | Monitoring, tuning, guardrail reviews and small fixes for 10 hours | From USD 2,500 per month |
| Model usage | Tokens consumed by conversations, lookups and generated output | Variable with volume |
| Platform licences | CRM, telephony and data tools the agent connects to | Set by each vendor |
| Human escalation | Staff time on cases the agent hands over | Internal cost |
| Knowledge upkeep | Refreshing articles and policies the agent answers from | Scoped as needed |
The recurring layer to plan alongside the build
Signal → Synthesis → System → Scale
Find where intelligence creates value.
Understand the business to identify and prioritize the opportunities with the greatest measurable impact.
Translate complexity into a clear design.
Bring together the people, workflows, data, and technology to define how intelligence should work.
Turn the design into a working capability.
Build the solution that embeds intelligence into how work and decisions happen.
Compound what works.
Measure impact, optimize performance, maintain reliability, and unlock greater leverage.
From signal to scale.
S4 frames AI agent spend as a fixed support retainer plus usage that scales with routed work, so finance teams can model monthly cost before approving a build.
| Step | Stage | What it changes |
|---|---|---|
| 01 | Signal | Log conversation volume, CRM and ticketing integrations, model usage charges, and the USD 2,500 monthly retainer baseline. |
| 02 | Synthesis | Set which workflows the agent handles, where handoffs to people occur, and how usage bands translate into predicted monthly spend. |
| 03 | System | Connect the company brain and CRM, test accuracy on routine questions, and instrument failure paths that increase support hours or compute. |
| 04 | Scale | Track cost per resolved conversation against retainer hours and usage, then extend the agent one workflow at a time. |
Ongoing costs start with a support retainer, which at Paloren begins at USD 2,500 per month for 10 hours. On top sit model usage charges that scale with conversation volume, plus any platform licences the agent relies on.
A small internal agent sits at the low end of that range, while an agent tied to a company brain and CRM needs more hours and compute each month.
Yes. The build is a one-off project, with AI agents at Paloren priced from USD 40k to 90k over 6 to 10 weeks. Once the agent passes testing and goes live, monthly support from USD 2,500 for 10 hours takes over, alongside usage and licence costs. Planning both layers together gives you the true first year figure.
Agents are custom software shaped around your data, systems and workflows, so a single monthly rate would hide the real work. Paloren quotes the build separately, USD 40k to 90k for agents over 6 to 10 weeks, then keeps the monthly layer transparent through a support retainer and visible usage costs.
That structure lets you see exactly what you are paying for each month.
The retainer scales to match the workload. Teams running several agents, or one agent handling complex integrations, often move to a larger monthly allocation so monitoring, tuning and escalation reviews are never rushed.
Paloren reviews usage with you so the hours follow the actual demand rather than a fixed assumption, and unused capacity can be redirected toward new workflows or additional agents.
Usually yes, at both layers. Voice agent builds run USD 25k to 60k over 4 to 8 weeks, above the USD 20k to 50k range for chatbots, because live calls need telephony integration and careful escalation design.
Monthly usage is also heavier since every call involves real time transcription and speech, and support hours lean toward call review and transcript analysis.
That is the usual path. Most businesses begin with a single agent on their highest value workflow, fund it with a build in the USD 40k to 90k range and a support retainer from USD 2,500 per month, then expand once the numbers hold.
Paloren plans the roadmap during strategy so later agents reuse the integrations and governance built the first time.
Support hours include reviewing conversations, escalations and task outcomes, and that review becomes reporting you can act on. You see where the agent handled work cleanly, where it escalated and which knowledge gaps caused errors.
Begin with an AI readiness assessment, from USD 8k over 2 to 3 weeks. It tests whether your data, systems and workflows can support an agent before you commit to a build or a monthly retainer.
If the answer is yes, an AI strategy engagement, USD 12k to 25k over 3 to 4 weeks, identifies the first workflow worth automating and the budget it justifies.
Paloren serves businesses worldwide, and engagement ranges are quoted at country level without offices or city surcharges. The build and retainer figures stay consistent, while usage costs vary with local volume, languages supported and any regional data residency requirements.
Those variables come out during scoping, so the monthly budget you approve reflects how the agent will actually run in your market.
What would an agent cost your team each month?
Book an AI readiness assessment