AI Agentic Systems Consultancy · Zurich

We Bring AI Into Your Business

Autonomous AI agents that do real work in your business, with governance built in from day one. We find where AI pays, build the agents and keep your people in control.

Example agent run

Order intake agentRunning
  1. Read the order PDF from orders@7 lines
  2. Mapped to your part numbers7 of 7
  3. Checked capacity per lineLine 4 over

Draft order confirmation

Waits for the scheduler to approve

ApproveEdit
Every step loggedKill switch readyRuns in Europe · Germany

One engagement, three layers

Audit, agentic system and governance, delivered as one

AI projects often stall between a slide deck and a compliance review. We run all three layers in one engagement, so what we find, what we build and how it is controlled stay connected.

01The audit

Find where AI pays

We map the part of your business that loses the most time, step by step, with your data and your systems. You get the agents worth building, ranked by effort and value.

1

Order intake

High value · Low effort

2

Delivery plan

High value · Medium effort

3

Invoice checks

Medium value · Low effort

How the audit works
02The agentic system

Build what runs

One agent on one process first, running next to your team on your real data. Then several agents, connected to the tools you already use.

Order PDFsYour ERP
Agent

Draft confirmation

Waits for approval

See what we build
03Governance, built in

Stay in control

Before any code, we make 144 decisions with you: what the agent may see, where your system runs, who approves what. Every action is logged and documented.

  • Approval before anything leaves
  • Every action logged
  • Kill switch on both sides
See the controls

The AI proposes, code executes, your people approve.

Agenticsis explained in 2 minutes

Subtitles
1:57 · English narration · subtitles in English, German and Spanish

What we build

Pick the part of your business that loses the most time

Every example is a starting point. In the discovery call we map your real process, your systems and your data, and build the version that fits.

2 examples for Operations.

Morning delivery plan

OperationsLogisticsManufacturingRetail and e-commerce
Example we can build

Every morning a planner rebuilds the day by hand from the order system, supplier emails and customer time windows. Late changes get missed until a truck has left.

Every weekday 05:30

What goes in

Order databaseyour ERP or order system
Supplier emailsdelays and changed quantities
Delivery windowsper customer
Fleet and capacityvehicles, drivers, shifts

What the agent does

  1. 1Collects new and changed orders since yesterday
  2. 2Reads supplier emails for delays
  3. 3Builds routes within capacity and time windows
  4. 4Marks every order at risk, with the reason

The planner approves or edits the plan

What comes out

Delivery plan, Tue 13 Octready 06:41
Orders planned48
Routes6
Within time window45
At risk: supplier delay3

In the planner's inbox before 07:00.

Usually starts asPilot
Time to first run4 to 6 weeks

Production order intake

OperationsManufacturing
Example we can build

Customer orders arrive as PDFs in different layouts. Planners retype them into the ERP and check capacity by hand before confirming a date.

New email in orders@

What goes in

Customer ordersPDF in any layout
Item masterpart numbers and routings
Capacity calendarmachines and shifts

What the agent does

  1. 1Reads the order and maps customer part numbers to yours
  2. 2Checks capacity for each line
  3. 3Proposes a delivery date
  4. 4Drafts the order confirmation

The scheduler confirms the date

What comes out

Order confirmation draftdrafted 10:05
Lines read7
Mapped to your parts7 of 7
Proposed delivery28 Oct
Line 4 exceeds capacitysplit delivery?

A draft order waits in your ERP for the scheduler.

Usually starts asPilot
Time to first run4 to 6 weeks

How we work

Four phases, and every phase ends with something you keep

After a free 45-minute call we start with discovery, test one agent on one process, and then you decide.

Phase 1

Discovery

1 to 2 weeks

  • Your process mapped step by step
  • A check of the data and systems involved
  • The agents worth building, ranked by effort and value

You keep: a written proposal for the next phase.

Gate G0: every decision recorded, risk classified

Phase 2

Pilot agent

3 to 6 weeks

  • One agent on one process
  • Built on your real data
  • Runs alongside your team, with approval steps

You keep: a working agent and a decision based on how it ran.

Gate G1: design and data routes approved

Phase 3

Production system

6 to 16 weeks

  • Several agents working together
  • Connected to your systems
  • Monitoring, handover and training

You keep: the system, its code and its documentation.

Gate G2: go-live review, two signatures

Ongoing

Care

Optional, cancel monthly

  • Monitoring and fixes
  • Model updates
  • Small extensions

AI usage is billed at cost, on its own line.

Gate G3 monthly review, G4 on every change

Your governance documents

Delivered with the system and updated whenever it changes.

  • Risk classification
  • Data map
  • AI register entry
  • System card
  • Impact assessment
  • Incident runbook
  • Contract annex
  • Jurisdiction annex

Our project bands next to a typical Swiss agency

CHF per project, log scale. Swiss agency ranges from our 2026 market research (20 sources).

1k5k20k80k250kSmallone workflow, portal or dashboard1.5k to 5kStandardseveral systems connected5k to 12kAdvancedAI assistant on your documents12k to 25kComplexmulti-channel, several agents25k to 80k
AgenticsisTypical Swiss agency

What shapes each project

We name these in the proposal, so the scope is clear from the start.

Systems to connectOne inbox is quicker than an ERP, a CRM and a shop together.
State of the dataClean exports go fast. Scanned PDFs and hand-kept spreadsheets take longer.
Your customers' languagesGerman, French, Italian and English answers each need testing.
Approval stepsWhere a person signs off, we build the screen for it.

Governance you can see

Built-in controls, not a policy on paper

Every system we deliver ships with the same controls. Their settings are decided with you in the first weeks and written into your documentation.

Where your system runs

You choose the server region at the start: Europe (Germany) or the Americas (USA). AI model processing under GDPR and Swiss FADP, with every provider listed on our Sub-processors page.

Americas · USA
Europe · Germany
Regions we serveSub-processors

144

decisions made with you before we build

From data access to who may stop the system. Nothing runs on a default you did not see.

Agent runningStop

A kill switch on both sides

Named people on your side and ours can stop the system at once. Who approves a restart is agreed per project and tested before go-live.

Your people approve what matters

A1Reads dataRuns alone, logged
A2Changes inside your systems, reversibleRuns, spot-checked
A3Sends something outsideOne approver
A4Irreversible or high valueTwo approvers

05:30Read 48 new ordersa3f9

05:31Drafted delivery plan7c1e

06:41Approved by the plannere04b

Every action logged

A hash-chained record of what each agent read, decided and did, kept for 12 months by default. You receive a signed fingerprint of it every day, so you can check it yourself.

T0Public
T1Internal
T2Confidential
T3Never sent to AI

Data sorted before it moves

Every data source gets a level. The most sensitive level is never sent to an AI model.

Built following EU AI Act requirements

Each use case is classified by risk before we build. Data processed under GDPR and Swiss FADP, and you receive the documentation.

GDPRSwiss FADPEU AI Act
Costs have a ceilingEvery run has a step, time and cost limit, plus a daily cap, so an agent cannot run away with your budget.
You own what we buildCode, workflows and documentation are handed over to you.
When we are not a fitIf a simple rule or an existing tool solves it, we tell you in the first call.

Integrations

Works with the tools you already use

We connect agents to more than 850 tools across CRM, communication, finance and operations. A few of them:

Microsoft OutlookMicrosoft TeamsMicrosoft ExcelSalesforceSlackand 850+ more
Founder's Perspective
“The companies that win with AI in the next three years won't be the ones with the biggest budgets. They'll be the ones that started with one specific problem and built from there. We call this AI mainstreaming: making the technology practical, accessible, and actually used.”
Sofía Salazar Mora, Founder & AI Systems Strategist, Agenticsis

Sofía Salazar Mora

Founder & AI Systems Strategist

Quick Answers

Direct answers to the questions clients ask before our first call.

Quick Answer

What is an AI agent, and how is it different from regular automation?

Regular automation follows fixed rules: if X happens, do Y. An AI agent reads context, makes decisions, and handles situations that were not pre-programmed. Example: a scheduling automation blocks time in your calendar. An AI scheduling agent reads the email, understands the urgency, checks your preferences, proposes the best slot, and sends the reply itself.

Quick Answer

How long does it take to deploy an AI agent or automation?

It starts with a free 45-minute call. Discovery then takes 1 to 2 weeks, a pilot agent on one process 3 to 6 weeks, and a production system with several connected agents 6 to 16 weeks. During the pilot you see the agent working in weekly demos, so you can redirect before we have built the wrong thing.

Quick Answer

What does the AI business audit actually deliver?

Discovery maps one process in 1 to 2 weeks: the steps, the data and systems involved, and the agents worth building, ranked by effort and value, with a written proposal for the pilot. The full AI Business Audit looks at the whole company across 8 readiness pillars and takes 2 to 8 weeks depending on size; its findings replace most of Discovery. Both include the risk classification of each use case under the EU AI Act.

Frequently Asked Questions

Everything you need to know about AI agents and Answer Engine Optimization.

How do I get started with AI automation for my business?

The easiest way is to book a free 45-minute call with us. We'll identify your biggest time-consuming tasks and show you exactly what could be automated, no jargon, no pitch deck. We recommend starting with one specific problem: a repetitive email process, a manual data-entry step, or a customer question that gets asked 50 times a week. You don't need an AI strategy or a technical team to begin.

How do I choose the right AI agency for my business?

Look for three things: (1) Do they start with your business problem or with technology? An agency that leads with 'we use GPT-4' is selling tools, not solutions. (2) Do they test one agent on one process with your real data first, rather than starting with a 6-month roadmap? (3) Can they support the system after deployment? AI systems need maintenance and adaptation. At Agenticsis, every project starts with a short Discovery and a pilot, every agent has a clear escalation path, and maintenance after handover is available as the optional Care phase.

Do I need to be a large company to work with Agenticsis?

No. We work with businesses of all sizes, from solo consultants to teams of 200. We recommend starting with a single automation. A first automation typically targets 5 to 10 hours of manual work per week, an estimate we check against your own process before we build. You don't need an AI strategy or a technical team to get started.

What is the difference between a simple automation and an AI agent?

A simple automation follows fixed rules: if an invoice arrives, file it in folder X. An AI agent can reason: it reads the invoice, identifies the supplier, checks for discrepancies against your records, and flags anything unusual for your review. Agents handle the grey areas that automation can't.

How quickly will I see results?

It starts with a free 45-minute call. Discovery then takes 1 to 2 weeks, a pilot agent on one process 3 to 6 weeks, and a production system with several connected agents 6 to 16 weeks. During the pilot you see the agent working in weekly demos, so you can redirect before we have built the wrong thing.

What tools and software do you integrate with?

We integrate with the tools your team already uses - HubSpot, Salesforce, Notion, Slack, Gmail, WhatsApp, WordPress, Shopify, and most European business software. If it has an API, we can connect to it. We build around your stack, not the other way around.

Do I need technical staff to maintain an AI system after deployment?

No. After handover your team uses the system, and we provide onboarding so your staff understands what it does and when to escalate to a human. If you want us to handle the upkeep, the optional Care phase covers monitoring, fixes and model updates.

Is my data safe when using Agenticsis AI agents?

Yes. We are based in Switzerland and every engagement follows the EU GDPR and the Swiss FADP. Data is stored securely and encrypted in transit and at rest, and we never sell client data. We share data only with the sub-processors listed on our Sub-processors page. Our security measures are listed in Annex II of our Data Processing Agreement.

Not sure where to start? That's exactly why we offer a free call.

We'll spend 45 minutes understanding your business. No pitch, no generic demos. Just an honest conversation about what's possible - and what's worth doing first.