Blog · Published · Updated
AI Agent for Freelancers: A Sourced Pre-Call Brief
Short answer
An AI agent for research before a call goes through agreed sources, separates facts from hypotheses and prepares opening questions, with a source for every conclusion. With 20 calls a month a freelancer gets back 10 hours, or €349, which a template in ChatGPT or Claude reaches more easily. Building an agent for €3,000 net pays back in a sales team with 100 calls in 1.7 months.
How we workDetails ↓
- 01Call30 minutes with an engineer, free
- 02PrototypeA working system on your process and data, free
- 03ImplementationTarget and fixed price in the contract; 50% at signing
- 04Acceptance50% once the agreed target is met
- net per agent for one process, build paid once
- from €3,000
- from signing the contract to acceptance
- 6–16 weeks
- free prototype on your cases
- €0
We belong to the OpenAI and Anthropic partner programmes and learn through their official courses. What this means ↓
Author: Syntalith · 3 min read
Statistics
AI in information work: statistics 2026
1.7 mo.
payback of a €3,000 agent in a team with 100 calls a month, when preparation drops from 45 to 15 minutes. For one person with 20 calls it is €349 a month. Scenario.
25.1%
According to a Harvard Business School study (2023) of 758 consultants, tasks within the model's capabilities went 25.1% faster with AI.
40%
According to a study published in Science (2023), ChatGPT cut the time 453 professionals spent on writing tasks by 40%.
Source: Noy and Zhang, Science (2023) ↗
32.7%
According to Eurostat (2025), 32.7% of people aged 16 to 74 in the EU used generative AI tools in the previous 3 months.
90%
According to the European Commission (2026), 90% of companies that use AI say people take the decisions, and 2% of decisions are fully automated.
Source: European Commission, Flash Eurobarometer on AI and algorithmic management (2026) ↗
Figures checked: 22 October 2026
What makes up an agent for client research
A call with a potential client in two hours: the company's website, the business register, recent news, the person's posts and your questions. The agent does part of this preparation and marks what it could not confirm.
Plan and tools
what to check every time
Search engine, websites, company registers, calendar, notes.
Your own materials
search in your offers
Case studies, call summaries, question lists for the industry.
Permissions and log
read-only to start
What the agent calls on its own, what waits for approval, a record of steps and costs.
A company description without sources is hard to check, so before the call you need access to the material the system based its conclusions on.
Which sources to allow
The client's website, public registers, press releases, job adverts and reviews marked as reviews. For a sole trader, a public entry in the business register still needs a purpose and a retention period when you process it further. LinkedIn is especially sensitive: its terms prohibit unauthorised scraping, so the agent only organises information you have checked yourself.
What it gives a freelancer, and what it gives a team
- Freelancer, 20 calls
- 20 × 30 min = 10 h (€349) a month
- Model, 8 calls per meeting, Claude Sonnet 5.5
- €3
- Sales team, 100 calls
- 100 × 30 min = 50 h (€1,745), model €16
A €3,000 agent pays back in the team in 1.7 months
The preparation and checking times are the scenario's assumptions. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat. A freelancer's rate is usually higher. Model: 6,000 input and 1,000 output tokens per call at Anthropic's rates, converted at the ECB rate of 1.1205 USD per euro on 22 October 2026.
For one person with 20 calls, a summary template in ChatGPT or Claude is more convenient than building an agent. According to a Harvard Business School study (2023), tasks within the model's capabilities went 25.1% faster with AI. Outside those capabilities the results got worse, so keep the assessment of the person and the decisions for yourself. The study on the jagged frontier of AI describes where that line runs.
What not to automate
- Sending messages to the client
- Publishing posts and comments
- Updating the CRM without approval
- Pulling data from platforms that prohibit automation
- Ratings such as „this lead is weak” without clear criteria
The agent starts as an assistant for reading and synthesis. Once the quality is stable, you can add a task in the CRM or a draft email, with manual approval for external actions. The NIST AI RMF describes risk management when you design and use such systems.
GDPR and control of data
A private individual's details in the summary are personal data, including the job title, a public profile and a sole trader's address. Set the purpose, limit the scope to preparing the call, record the sources and the date, do not build a permanent profile and set a retention period. „We do not train on customer data” does not mean the provider stores nothing. We build an agent for a sales team with limited permissions, sources and approval of actions as part of AI agents. For the free process review, count the calls and the real preparation time from the last month.
