In short
An AI agency for SMEs takes on four jobs: strategy, development, rollout in live operation and training the team. Pharen IT is a GmbH from Naunhof near Leipzig that offers exactly these four services, on site in Leipzig and Saxony or remote. The starting point is the Automation Sprint: in four weeks, one real process such as incoming invoices or lead qualification is automated with AI, with approvals, a pilot run and an evaluation with before and after values.
Pharen combines consulting with a product of its own. Pharen Hub is an Open Fair-Code AI workspace that runs in the managed cloud or on your own infrastructure. The business team needs about two to four hours per week during the sprint.
Two to four hours per week. That is how much time your business team needs in our Automation Sprint, and we take care of the build. The number comes first because it answers a question that stands before every project: an AI agency for SMEs should take work off your hands and not create another project next to the daily business.
This guide describes what an AI agency should deliver, how working with us runs and which questions you can use to check any vendor. We are a vendor ourselves: Pharen IT is a GmbH based in Naunhof near Leipzig (Amtsgericht Leipzig, HRB 44633).
What does an AI agency for SMEs do?
An AI agency brings AI into real processes of a company. At Pharen that means four services that build on each other.
| Service | What happens | Result |
|---|---|---|
| Strategy | We determine where AI should support, what stays with people and which rules apply to data, roles and decisions. Benefit, feasibility and risk are assessed together. | A first process with an owner and clear success criteria |
| Development | We connect data, roles, approvals and integrations and build the use case as a testable workflow with AI agents. | A working process in your workspace |
| Rollout in operation | Your users work with real cases, in parallel to the existing way. Exceptions are collected, recognition and rules are corrected. | A log of the pilot cases and an evaluation with before and after values |
| Training | Documented processes, clear roles and short training sessions enable the owners to adapt and extend agents themselves. | A team that keeps working without lasting dependence on external consultants |
The overview of all services is on the services page. For teams in Leipzig and Saxony there is the page on AI consulting in Leipzig.
How does a project with an AI agency run?
The starting point with us is the Automation Sprint. It takes four weeks, and every week has a fixed result.
- Week 1, selection and process map. We check two to three candidates against business goal, volume, data access and ownership and decide on one process together. The result is a one-page process map and a baseline measurement, for example cycle time and number of open cases.
- Week 2, build. We build the process in Pharen Hub: intake, automatic preparation of the data, check for missing details, status and approval step. Every action that touches money, customers or contracts stays tied to a human approval. Before the first case, your IT or your data protection team confirms the operating model, data paths and approval steps.
- Week 3, pilot run. The team handles real cases through the new process. On day 3, 7 and 10 we look at the cases together and record where people had to step in.
- Week 4, evaluation. We compare the pilot numbers with the baseline measurement. That leads to a recommendation: expand, rework or stop cleanly.
The sprint is not a strategy workshop and not a software rollout. It fits teams with one specific process that runs today by mail, spreadsheet and word of mouth.
Which processes are good to start with?
We have documented two processes publicly.
Incoming invoices. Invoices arrived as a mail attachment, an upload or a forwarded message, and follow-up questions sat in separate threads. The new process has one intake for mail and upload, recognises supplier, amount and due date, checks for missing details and leads to approval with responsibility and history. It is measured by the time to first review, the time to approval, overdue invoices and manual corrections. The process is described under incoming invoice process.
Lead qualification. Leads came in on the website, by mail, through referrals, LinkedIn and conversations. The new process has one intake for form, mail and manual contacts, a simple qualification by need, relevance and timing, routing to the right person and follow-up reminders. The process is described under lead generation.
Other candidates are onboarding and order review. More on the method is under AI process automation.
How do you recognise a good AI agency?
You can put five questions to any vendor, including us.
- Does the agency start with a process of yours or with a tool of its own?
- Where is the data stored, and which model processes it? Get the DPA before the first test.
- Which steps stay tied to a human approval?
- How is success measured, and is there a baseline measurement?
- Can your team run and extend the process itself after the project?
We name one limit openly. We do not publish blanket savings figures. Reliable values only emerge with a defined starting point and a comparison period that is long enough. We do not currently list an ISO 27001 certification. What we do technically and organisationally is on the security page.
What sets Pharen apart from a pure consultancy?
Pharen combines consulting with a product of its own. Pharen Hub is an Open Fair-Code AI workspace with team chat, projects, documents, workflows and AI agents. It runs as a managed cloud or on your own infrastructure. So the automated process does not sit in a script that only the agency understands. It sits in a workspace where your team sees status, sources and results. The guide to the on-premise AI platform describes what operation on your own servers looks like.
If you want to prepare AI internally first, the guide KI im Unternehmen einführen (German) comes with a PDF and an Excel template.
Frequently asked questions
What does an AI agency cost for a mid-sized company?
The price depends on the process, the integrations and the operating model. In the first call we clarify the goal, the tools involved and the most important approval point, and after that the effort can be named. The prices for Pharen Hub itself are on the pricing page: Free, Starter €25, Team €79 and Pro €199 per workspace and month, self hosted licence from €20.
How long does it take to automate a process with AI?
In the Automation Sprint, four weeks: selection and process map, build, pilot run with real cases and evaluation. The business team needs about two to four hours per week for it.
Does Pharen only work in Leipzig and Saxony?
No. Pharen IT is based in Naunhof near Leipzig and advises teams on site in Leipzig and the region as well as remote.
Does Pharen also offer AI training?
Yes. Training is part of every project: the owners learn to adapt agents and workflows themselves. On top of that there are the guides and learning paths of the Academy.
Does our data stay in Germany?
You define that in the project. Pharen Hub runs in the managed cloud or on your own infrastructure, and you decide which models are connected. Operating model, data paths and approval steps are confirmed by your IT or your data protection team before the first case.
So two to four hours per week over four weeks are enough to know whether AI holds up on one of your processes. If you have a process in mind, talk to us.
