Humanika: the new AI initiative by 150 Por Ciento
150 Por Ciento logo 150 Por Ciento logo

AI agent development for businesses

An AI agent is software that works toward a goal by taking actions through tools — updating your CRM, checking an order in the ERP, opening a helpdesk ticket or replying on WhatsApp — within the rules and human approvals you define. At 150 Por Ciento we design, build and operate agents connected to your systems and documents, with evaluation and governance from the start. Projects start at 3,000 USD.

AI agent development for businesses

Technologies we use

  • Claude API
  • OpenAI API
  • Model Context Protocol (MCP)
  • LangChain
  • RAG
  • pgvector
  • WhatsApp Business API
  • Python
  • Node.js
Starting from USD 3,000

In short

AI agent development is the design and construction of AI systems that use tools to act on a company's CRM, ERP, helpdesk, documents and WhatsApp, with guardrails, human approval for sensitive actions and continuous evaluation.

What we solve with AI agents

Most teams don’t need another chat window. They need repetitive, multi-step work to get done without someone copying data between systems. Typical cases:

  • Sales reps updating the CRM by hand after every call or WhatsApp conversation
  • Back-office staff checking orders, invoices or inventory across the ERP and email
  • Helpdesk teams triaging, classifying and routing tickets that follow known rules
  • Analysts searching contracts, manuals and policies to answer internal questions
  • Pilots that worked in a demo but never got connected to real systems

An agent is worth building when the task has clear inputs, repeatable steps and an outcome you can verify. If it doesn’t, we tell you before you spend the budget.

What an AI agent is — and how it differs from a chatbot or applied AI

An AI agent is a language model that receives a goal, decides which tools to use and executes actions in your systems, checking results along the way. The difference with a chatbot is not the interface but the permission to act.

AI chatbot AI agent Applied AI
What it is A conversation channel (WhatsApp, web) Software that executes tasks with tools The strategy and roadmap that decides where AI goes
Main output Answers and handoffs to a person Actions in CRM, ERP, helpdesk, documents Prioritized use cases, architecture and governance
Typical risk Wrong or invented answers Wrong actions with real consequences Investing in the wrong use case
At 150 Por Ciento AI chatbots for WhatsApp and web This service Applied AI & AI agents

The label matters less than the substance. In June 2025 Gartner predicted that over 40% of agentic AI projects will be canceled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls, and warned about “agent washing”: chatbots and RPA rebranded as agents (Gartner, 2025). That is why we start with one measurable process, not with a platform.

What our agents connect to: tools, RAG and MCP

An agent is only as useful as the systems it can reach. We build three layers:

  • Tool use: typed functions the agent can call — create a lead, check stock, open a ticket, send a WhatsApp template — each with validated inputs and the minimum permissions required.
  • RAG over your documents: retrieval-augmented generation over contracts, manuals, policies and knowledge bases, so the agent reasons with your information and cites the source.
  • MCP integrations: the Model Context Protocol, released by Anthropic in November 2024 as an open standard for connecting AI assistants to the systems where data lives (Anthropic, 2024). In December 2025 it was donated to the Agentic AI Foundation under the Linux Foundation, with more than 10,000 active public MCP servers reported at the time (Anthropic, 2025). Building on MCP keeps your tools reusable across models and vendors.
System Example actions
CRM (HubSpot, Salesforce, others) Log calls, update deal stages, create follow-up tasks
ERP Check orders, inventory and invoice status; prepare drafts for approval
Helpdesk Classify, prioritize and route tickets; draft replies with sources
WhatsApp Business API Receive requests, confirm data, send approved templates
Documents and intranet Search, compare and summarize with citations

Evaluation, guardrails, human-in-the-loop and governance

An agent that acts needs more control than one that only talks. Our standard setup:

  • Evaluation before every release: a test set built from your real cases, measuring task success, wrong actions and escalations.
  • Guardrails: least-privilege tools, input and output validation, protection against prompt injection and limits on cost and number of steps.
  • Human-in-the-loop: high-impact actions — payments, contract changes, messages to key customers — wait for a person’s approval.
  • Traceability: every decision, tool call and source is logged and auditable.
  • Governance: owners, permissions and review cycles aligned with frameworks such as the NIST AI Risk Management Framework (2023).

This is not theoretical: OWASP lists “Excessive Agency” — agents with too many tools, permissions or autonomy — among the top 10 risks for LLM applications (#3 in the 2026 edition) and recommends human approval for high-impact actions (OWASP, 2025). Security reviews can be extended with our cybersecurity team.

Processor chip on a digital circuit board

How we work: from pilot to production

  1. Discovery: we map the process, the systems involved and what the agent should and shouldn’t do. You receive a proposal within 24–48 hours after the call.
  2. Pilot: one process, a limited set of tools, real data and human approval on sensitive actions.
  3. Evaluation: we measure the pilot against the baseline and decide with you whether to scale, adjust or stop.
  4. Production and operation: monitoring, new tools added gradually and periodic review of permissions.

Development projects start at 3,000 USD. We apply the same practices in Humanika, 150 Por Ciento’s own AI initiative of digital collaborators that work on web and WhatsApp and connect to enterprise systems — it is where we prove the approach before recommending it. Behind it is a software, AI and app development agency founded in 2015, with offices in Bogotá and Lima and more than 200 projects launched. Tell us about your process and we’ll reply within 24 business hours.

Frequently asked questions

What is the difference between an AI agent and an AI chatbot?




A chatbot is a conversation channel: it answers questions on WhatsApp or your website. An AI agent takes actions: it calls tools to read and write in your CRM, ERP or helpdesk to complete a task, inside limits and approvals you define. Many projects combine both: the chatbot is the front door and the agent does the work behind it.

How much does it cost to build an AI agent?




Development projects at 150 Por Ciento start at 3,000 USD. The final cost depends on how many systems the agent connects to, how many actions it can take and how much evaluation and governance the process requires. Model API usage is billed separately, is variable and is reported transparently.

How long does an AI agent pilot take?




It depends on the process, how many systems the agent touches and how ready your data is. A first pilot is limited to one process, a small set of tools, real data and human approval on sensitive actions, and its timeline is confirmed in the proposal we send within 24–48 hours after the discovery call.

Can an AI agent connect to our CRM, ERP and WhatsApp?




Yes. Agents use tool calling through each system’s API, and where it fits we expose those tools through the Model Context Protocol (MCP), an open standard for connecting AI applications to business systems. Each tool gets the minimum permissions it needs, under the agent’s own credentials.

How do you stop an AI agent from taking the wrong action?




With layers: least-privilege tools, validation of every action before it runs, human approval for high-impact steps, full logging and an evaluation suite the agent must pass before each release. Excessive permissions and autonomy are a recognized risk in OWASP’s Top 10 for LLM applications (#3 in the 2026 edition), so we design against it from day one.

Does 150 Por Ciento build its own AI agents?




Yes. Humanika is 150 Por Ciento’s own AI initiative: digital collaborators that work on web, WhatsApp and other channels and connect to enterprise systems. It is where we test in production the architecture, evaluation and governance practices we apply in client projects.

Read our Privacy Policy to learn how we protect your data and respect your privacy.