From Email Chaos to AI Agents: Highlights from the Visagio + Pipefy Webinar

From Email Chaos to AI Agents: Highlights from the Visagio + Pipefy Webinar

In our recent webinar, Visagio and Pipefy teamed up to share how AI-powered, no-code workflows are transforming back-office operations—starting with very real, very messy processes like HR service desks and expense reimbursements.

Who’s Who: Visagio & Pipefy in a Nutshell

Visagio

• Started in Brazil in 2003 as a tech-driven management consultancy.

• Today: ~1,000 people globally, with over 130 consultants in Australia (Perth, Sydney, Brisbane and presence in Melbourne).

• Strong focus on operational excellence, process improvement, supply chain, and digital/AI solutions.

• AI-first since day one: early work included optimization projects (e.g. sports scheduling), and they later launched an AI academy (ViAcademy) to train new AI talent and feed skills back into the market.

Pipefy

• A 10-year-old company with 4,000+ customers in 150+ countries.

• No-code/low-code workflow and process orchestration platform.

• Strong partner ecosystem, including Visagio, to deliver local, high-impact implementations (e.g. Australia, India).

• Currently evolving from no-code to natural language – enabling business teams to configure workflows and AI agents by simply describing what they need.

Why This Is a “Revolution” (Not Just Another Tool)

The speakers framed the current moment very clearly:

• By 2026, the vast majority of enterprise companies are expected to be running generative AI in production.

• Over 80% of large-company CEOs expect GenAI to impact business value in 2024–2025, not in some distant future.

But despite the hype, there are five very real barriers that keep many companies stuck in pilots:

1. Technical implementation – vectorization, tokenization, model adaptation, infra, etc.

2. Token and infrastructure cost – running large models at scale.

3. Talent scarcity – AI expertise is usually concentrated in IT/Data, not in HR, Finance, Legal, etc.

4. Data governance – security, compliance, and responsible AI.

5. Data availability and quality – training models on messy, fragmented operational data.

Pipefy’s entire value proposition is essentially built around these pain points:

“We were born to enable non-technical teams to do what they do best, better.”

From No-Code to Natural Language Orchestration

Pipefy positions itself as an orchestration layer for processes and data:

• Connects legacy systems, ERPs, databases, and virtually anything with an open API.

• Orchestrates data from multiple sources (forms, portals, email, chatbots) into a single workflow.

• Automates manual work: moving records, assigning owners, sending emails, running approvals, calling integrations, and now… invoking AI agents.

The next step in this evolution:

• A natural language workflow builder (“I want to build a reimbursement process based on finance best practices…”).

• An AI agent library with preconfigured agents for common use cases (e.g., recruitment, invoice analysis, procurement, supplier onboarding).

• A no-code prompt builder to define how each agent behaves, using fixed and dynamic fields—no need for deep AI engineering.

Case Study #1: Gold Mining Company – HR Service Desk Transformation

Context

A large gold mining company in Australia had an HR service delivery team relying on… a shared mailbox:

• 16-person HR hub team.

• Over 3,000 emails per month.

• More than 2,000 manually signed documents per year.

• 12 categories and 63 different HR processes (onboarding, offboarding, org changes, leave queries, etc.).

• Tickets were emails, manually classified, manually tracked, often lost or delayed.

The original client request was modest:

“Can you build a script to count how many tickets/emails we get per month?”

Visagio stepped back and reframed the challenge: the real problem wasn’t reporting—it was the lack of structure, visibility, and automation.

What They Implemented

1. Discovery & Design

   • Interviews with stakeholders across HR and operations.

   • Full mapping of the 63 processes, rationalizing and simplifying into 38 optimized processes.

   • Selection of Pipefy as the best-fit platform (after evaluating multiple tools).

2. Solution Architecture

   • A central portal where all 3,500+ employees can submit HR requests via structured forms.

   • Automations to:

     • Classify and route tickets.

     • Assign owners.

     • Trigger emails and DocuSign flows.

     • Move cards between phases based on rules.

   • Integrations with:

     • SAP SuccessFactors (people and position data).

     • DocuSign (for contract and document signatures).

   • Reporting layer connected to Pipefy data to monitor volumes, SLAs, and performance.

3. Ticket Management Board

   • A consolidated board where the 16 HR hub members work from one single view:

     • All tickets from all sites.

     • Status, SLA, and assignee.

     • Late/at-risk items easily visible.

4. Change Management

   • Training sessions and how-to videos with voiceover.

   • Clear deployment checklists and support for on-site and office users.

Results (12-Month Period)

• 2,306 tickets processed through Pipefy.

• 840,000+ automation jobs executed.

• 45,000+ automated emails sent.

• 30,000+ DocuSign actions triggered.

• Annualized savings: ~AUD $4 million (time saved from manual email handling, rework, and chasing).

The key win: HR now works from a structured, transparent, and automated system instead of an overflowing shared inbox.

Case Study #2: Visagio’s Own Reimbursement Workflow + AI Agent

The second case was “eating their own dog food”: Visagio redesigned its expense reimbursement process using Pipefy and an AI agent.

Before Pipefy

• Reimbursements were requested using an Excel template.

• Users had to:

  • Download the spreadsheet.

  • Fill in free-text fields (plenty of room for errors).

  • Attach invoices and send everything to Finance.

• High levels of rework and inefficiency:

  • Wrong cost types and values.

  • Missing or incorrect client info.

  • Poor onboarding for new joiners.

• 19%+ of cases required rework/support from Finance.

After Pipefy + AI

1. Streamlined Request Form

   • Web and mobile-friendly.

   • Mandatory fields for:

     • Requester, approver, cost center.

     • Expense date, expense type (drop-down), amount.

     • Invoice attachment.

   • Requests can be submitted in seconds, not minutes.

2. Automated Communication

   • Requester receives a confirmation email with all details.

   • Approver and Finance receive structured notifications with links to the card and invoice.

3. AI Reimbursement Agent

   • Configured directly inside Pipefy with a natural language prompt.

   • The agent:

     • Reads the invoice.

     • Compares it with the reimbursement data.

     • Checks purchase date, values, and classification.

     • Produces a short analysis and recommendation (e.g., “All details match; safe to approve.”).

   • The AI can be configured to:

     • Auto-approve/deny based on rules, or

     • Suggest a decision, leaving the final call to a human (Visagio chose this human-in-the-loop model).

4. Approver Experience

   • Approver receives an email with:

     • Request and cost details.

     • Invoice link.

     • AI agent’s analysis and recommendation.

   • One click opens the Pipefy card for final review and approval/denial.

   • Can also manage all pending approvals directly inside Pipefy (no need to search emails).

5. Finance Team Experience

   • Finance sees:

     • Full request history.

     • All AI analysis and comments.

     • All emails and approvals in the card timeline.

   • They complete a simple checklist and finalize the payment.

   • Integration with the financial system is the next step, to reach full end-to-end automation.

Results

• Implementation time: ≈3–4 weeks from mapping to go-live.

• <1.2% rework (vs. >19% previously).

• 2,000+ AI analyses supporting invoice validation.

• 800+ reimbursement tickets processed.

• 13,700+ automation jobs and 5,000+ automated emails executed.

The process is now simple for users, efficient for Finance, and auditable and measurable for leadership.

Q&A: What the Audience Wanted to Know

1. How long does it take to go from kickoff to production for a midsize workflow?
Typically 1–4 weeks, depending on:

• Process complexity.

• Number of integrations.

• Level of AI/automation required.

The reimbursement process was done in about 3–4 weeks; some simpler workflows can go live faster.

2. Which AI models does Pipefy support?

• Out of the box, Pipefy uses OpenAI (GPT models).

• You can also plug in:

  • Gemini, Azure OpenAI, or other major providers.

  • Your own corporate AI endpoint (BYOM – bring your own model).

Visagio’s internal use case currently runs on the default GPT-based setup.

3. How is pricing structured?

• Licensing is per user (i.e., per operator actually working inside the pipes).

• Requesters and approvers interacting via email forms/portals typically do not count as licensed users.

• For companies with 100+ employees, the Unlimited plan is usually recommended:

  • Unlimited automations.

  • Generous API limits.

  • AI credits included.

4. What happens when the AI misclassifies or cannot read the data?

• You can explicitly instruct the agent:

  • If the invoice quality is low or fields can’t be reliably extracted, the agent flags the case.

  • It recommends a manual review by the Finance team rather than giving a false “approved” recommendation.

• This is fully configurable in the prompt and rules, so you keep control over risk.

5. Can this work for complex, value-creating processes—not just standard admin ones?

Yes—but mapping is critical:

• For higher-risk or complex processes (e.g., safety-related checks, specialized procurement), more effort goes into:

  • Detailed process mapping.

  • Defining decision logic and guardrails.

  • Pinpointing where humans must stay in the loop.

Examples discussed:

• Procurement of PPE equipment, where SKU structures are well defined: mapping + Pipefy build can be done in about a week.

• More complex processes may take closer to a month from mapping through agent configuration.

6. How does Pipefy integrate with Microsoft Copilot, Google, or other AI ecosystems?

• Pipefy can integrate with your existing AI stack.

• You can:

  • Use Pipefy’s native AI.

  • Or route requests to Copilot, Google AI, or internal models to respect data governance and security policies.

In other words, Pipefy becomes the process and orchestration layer, not a competing AI silo.

Key Takeaways

• You don’t need to start with moonshot AI projects. Start where:

  • There’s heavy manual work,

  • Relatively low risk, and

  • Clear, measurable outcomes (e.g., HR tickets, reimbursements, procurement, AP/AR).

• The combination of:

  • Structured workflows (Pipefy),

  • Process excellence (Visagio), and

  • Configurable AI agents
    can deliver very tangible value in weeks—not years.

• AI doesn’t have to replace human judgment. A human-in-the-loop setup often strikes the best balance between efficiency, control, and trust.