Agentic AI & Automation · Vienna
From automation to autonomy.
We design, deploy and govern the AI agents that run enterprise operations — from RPA foundations to autonomous workflows, engineered for regulated industries in Austria and Germany.
The shift
Most enterprises have automated the tasks. The work still belongs to people — the reading, the deciding, the chasing across systems. Agents change that: they take ownership of outcomes, act across your systems and stop when they should — within limits you set and can audit. BOTFORCE designs, builds and governs them.
What we build
Four service lines. One governed agent workforce.
From the first opportunity map to a running fleet — every step leaves you with agents that have an identity, a scope and an audit trail.
-
01
Agent Strategy & Readiness
Where agents create value in your operations — and where they must not.
We map your processes for agent potential, score them by value, risk and data readiness, and design the target architecture on top of what you already run — UiPath, SAP, Microsoft 365, your own systems. You leave with a business case your CFO signs and a roadmap your CISO accepts.
-
02
Agent Engineering
Agents that read, reason and act inside your real systems.
We design and build the agents: the tool layer that lets them post to SAP, drive a UiPath robot or call your APIs; the evaluation suite that proves they decide correctly; the human-in-the-loop workflows that catch the rest. Pilots go live in weeks and are built for production from the first commit.
-
03
Agent Governance & Identity
Every agent gets an identity, a scope, an audit trail — and a human who can stop it.
Agents are new principals in your access model. We give each one a first-class identity in Okta or Entra ID, least-privilege scopes with short-lived tokens, guardrails as code and an immutable audit trail — and we map the whole stack to the EU AI Act so transparency, oversight and documentation duties are met by design.
-
04
Agent Operations
We run the fleet so your team runs the business.
Monitoring, exception queues, cost control and continuous evaluation — delivered from one governed operations hub with an isolated tenant for your company. Models, prompts and policies are reviewed monthly; every change is versioned and every decision stays replayable for your auditors.
How agents work
The anatomy of a governed agent.
Five moves in a loop, wrapped in four controls. Hover a stage — or wait, and the agent walks you through it.
The loop
The controls
The path
From automation to autonomy, one governed step at a time.
Most operations sit at level two. Agents take you to four — without skipping the controls that make it safe.
-
L0
Manual work
People move data between systems by hand. Knowledge lives in inboxes and in the heads of the two colleagues who know how it really works.
-
L1
Scripted automation
RPA bots replay fixed steps. Fast and reliable — until the invoice looks different, the screen changes or an exception appears.
-
L2
Intelligent automation
Document AI and ML decisions inside the flow. Straight-through rates go up; every exception still lands on a person's desk.
Most enterprises today -
L3
Assisted agents
Agents read, reason and propose complete outcomes. A human approves every action — the trust-building phase.
-
L4
Governed autonomy
Agents act within identity and policy limits. Humans supervise by exception, with a full audit trail and a stop switch.
Where BOTFORCE takes you
Where agents work
Real functions. Real systems. Human limits.
Six places where governed agents take over the work — and where the human stays in the loop.
The accounts-payable agent
Finance · Shared services
Takes every incoming invoice from mailbox, EDI or portal, extracts and validates the data, runs the three-way match against purchase order and goods receipt, proposes the posting and the payment run — and answers the vendor's "where is my money" e-mail with the real status.
- Systems
- SAP S/4HANA FI · MMUiPathDMSE-mail · EDI
- Guardrails
Posting limits per vendor class, vendor master whitelist, no bank-data change without human confirmation, four-eyes above threshold.
- Human in the loop
Exceptions land in a queue with an SLA; the agent explains why it stopped and what it would do next.
The sourcing agent
Procurement · Supply chain
Turns approved requests into purchase orders, runs supplier onboarding checks — registry, sanctions, certificates — extracts terms and discount windows from contracts and keeps them in front of the buyer before they expire.
- Systems
- SAP MM · AribaSupplier portalsContract DMSRegistry APIs
- Guardrails
Budget and approval matrices as code, single-source purchases flagged, no new supplier without a completed check file.
- Human in the loop
Buyers approve supplier activations and any order above their delegated limit.
The onboarding agent
HR · People operations
Orchestrates a new hire's first day across HR, IT and facilities: accounts, equipment, trainings, documents in the right language — and answers policy questions from the handbook, with the source paragraph attached.
- Systems
- SAP SuccessFactorsEntra ID · OktaMicrosoft 365Ticketing
- Guardrails
Data minimisation for personal data, role-based access only, works-council agreements encoded as policy, no salary data in prompts.
- Human in the loop
HR confirms contract data once; managers approve access beyond the standard role profile.
The service-desk agent
IT operations · Security
Triages tickets, fulfils access requests against the policy model, runs incident runbooks step by step and writes the change documentation nobody else wants to write — with every action traceable to a ticket.
- Systems
- ServiceNow · JiraOkta · Entra IDMonitoringCMDB
- Guardrails
Change windows enforced, no production change without an approved change record, privileged actions time-boxed.
- Human in the loop
On-call engineers approve any step marked destructive; the agent prepares the rollback first.
The compliance agent
Compliance · Legal · Risk
Keeps the AI system inventory current, checks that agent-generated outputs carry the required transparency marks, collects evidence for audits and detects when a running agent drifts away from its approved policy.
- Systems
- Model registryAudit logGRC toolsPolicy engine
- Guardrails
Read-only by default, every finding linked to evidence, no external filing or notification without sign-off.
- Human in the loop
Compliance officers sign every filing; the agent drafts, references and reminds.
The tender agent
Public sector · Bid management
Monitors Austrian and EU tender portals, scores opportunities against your capability profile, drafts the response skeleton from past bids and checks every formal requirement before a human ever opens the document.
- Systems
- Tender portals · TEDDMSCRMPast-bid library
- Guardrails
Every claim cited to a source document, no submission without human sign-off, deadlines mirrored to the calendar.
- Human in the loop
Bid managers decide go/no-go on a one-page scorecard the agent prepares.
Governance
Autonomy you can audit.
The EU AI Act sets the deadlines. Our governance stack makes meeting them routine — and makes your agents trustworthy in the meantime.
-
One identity per agent
Agents are first-class principals in Okta or Entra ID — never shared service accounts. Every action carries the agent's name.
-
Least privilege, short-lived
Scoped tokens with a TTL of minutes, not standing rights. What an agent may touch and how much it may move is policy, not habit.
-
Immutable audit trail
Every perception, decision and action is logged and hash-chained — replayable for internal audit, the regulator or your own root-cause analysis.
-
Human oversight by design
Exception queues, four-eyes approvals and a stop switch for every agent — Article 14 built in, not bolted on.
Start with the readiness assessment
Fixed scope · effort quoted in person-days
-
2 Aug 2025
General-purpose AI obligations apply
Transparency and documentation duties for the foundation models your agents are built on.
-
2 Aug 2026
Article 50 transparency obligations apply
People must be told when they interact with an AI system; AI-generated content must be marked.
-
2 Dec 2026
End of the marking grace period (Art. 50 (2))
Systems placed on the market before 2 August 2026 must mark their outputs from this date.
-
2 Dec 2027
High-risk obligations — Annex III
Stand-alone high-risk systems: employment, credit, essential services, law enforcement and more.
-
2 Aug 2028
High-risk obligations — Annex I
AI embedded in regulated products such as machinery and medical devices.
Dates reflect the AI Act as amended by the Digital Omnibus. Not legal advice — we bring the lawyers to the table when needed.
Operations
One hub. Every agent accountable.
Hub-and-tenant: one governed operations hub, an isolated tenant per client, humans in every exception path.
Fleet monitoring
Every agent, its identity, its scope and its current task on one screen — per tenant, with no data crossing tenant lines.
Exception queues with SLAs
When an agent stops, a human sees why, what it would do next and how long the case has waited.
Cost and model governance
Spend per agent and per outcome, model and prompt versions under change control, evaluations rerun before every release.
Agent Operations Hub · Tenant view
Illustrative
Agents running
18
Waiting for human
3
Tasks today
2,418
Cost today
€41.20
| Agent | Tenant | Status | Last action | p95 |
|---|---|---|---|---|
| Invoice agent | Bank · AT | Running | 3-way match → posting proposed | 0.8 s |
| Compliance agent | Bank · AT | Running | Art. 50 output check · pass | 1.2 s |
| Onboarding agent | Med-tech · EMEA | Waiting for human | Access beyond role profile | — |
| Tender agent | Public · AT | Running | 6 tenders scored · 1 go | 4.1 s |
| Sourcing agent | Med-tech · EMEA | Paused · review | New supplier check file | — |
Sample data for illustration · identities, scopes and audit trail per tenant
Delivery record
Built by people who have run the go-lives.
Agents change how work gets done. Getting them into a regulated enterprise is still a delivery problem — the kind we have solved for 25 years.
Banking · Austria
Delivery management for a credit-claims value stream at a leading Austrian bank.
Regulated environment, several squads, hard regulatory deadlines — the discipline our agent rollouts inherit.
Med-tech · EMEA
EMEA cutover lead for a global manufacturer's S/4HANA programme.
Five-country go-live across Italy, Spain, Portugal, Turkey and South Africa, run from a live cutover dashboard we built ourselves.
Product · In-house
20+ products shipped by our own team.
From tender intelligence to planning tools — we build and run software, not slides. Our agents are built the same way.
“Autonomy without governance is a liability. Governance without autonomy is a spreadsheet. We build the part in between — agents that do the work and can prove it.”
Robert Aksan
Founder & Managing Director, BOTFORCE GmbH
- CredentialsDI (FH) · PMP · Zertifizierter KI-Manager
- Experience25+ years of enterprise delivery — SAP S/4HANA programmes, cutover and go-live management, RPA and UiPath automation
- TeachingPMI Authorized Training Partner · lectures in project management
- BaseVienna — Twin Towers, Wienerbergstraße 11/12A
Frontier notes
What we are watching.
Short, opinionated notes on the moves that matter for agentic operations in the DACH region.
Agent identity just became a product.
What Okta's Agent SSO and cross-app access mean for the way you grant agents access — and why the shared service account has to go.
Read the noteArticle 50, in practice.
How marking and transparency obligations apply to agent-generated documents, e-mails and decisions — and what to fix before 2 December 2026.
Read the noteFrom RPA centre of excellence to agent operations.
Why the hub-and-tenant model replaces the bot farm — and what your automation team does next.
Read the note
Start
Give your operations an agent workforce — governed from day one.
Book a strategy briefing
45 minutes · remote or in Vienna
Company
BOTFORCE GmbH Twin Towers · Wienerbergstraße 11/12A1100 Vienna, Austria
Write to us
robert@botforce.atPartnerships
UiPath Silver Partner · PMI Authorized Training Partner