Conflux logo - Evidence-based design - UX Research and Design Agency Milan Rome
Back to all solutions

Agentic Service Design

In Conflux Let's design AI agents tailor-made that really work, because they arise from business processes and from the real needs of users.

We apply Service Design and Agentic UX to create AI solutions that bring a measurable ROI and are actually adopted by teams, transforming AI from a technological “add-on” to an integral part of real processes.

The future of experience will no longer be designed in pixels or static wireframes, but designed as a reliable collaboration between people and intelligent systems. This is why we do not create rigid interfaces, but dynamic systems: experiences that evolve over time, adapt to real contexts and remain alive with those who use them.

To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
To truly shine, you've got to radiate brilliance. Unleash your imaginative flair with Steezy.
How do we guide the analysis of
Customer Experience
for companies.

We design an AI agent that is useful, usable and adoptable, ensuring that the solution has a measurable impact on business KPIs, such as reducing cycle times (MTTR) or increasing internal NPS.

If you don't design the experience around the agent you risk:

$2.1M
Burned on average for each AI project that doesn't arrive in production
Source: McKinsey*
42%
of AI projects abandoned in 2025, compared to 17% in 2024
Source: Gartner, 2025
32%
It crashes exactly after the pilot — not before, after
Source: MIT Sloan
40%
of the computational budget wasted just choosing which tool to use, with 50+ tools available
Conflux Search

A chatbot answers.

An AI agent collaborates.

A traditional chatbot follows scripts: it recognizes keywords and returns predefined answers. It works as long as the context is simple and predictable but as soon as the request leaves the expected path, it merely deviates or is not useful.

‍
It's a superficial tool, giving the illusion of support without really solving the problem.

An AI agent, on the other hand, interprets the context, makes decisions and acts within processes: it can access systems, manage exceptions and complete complex tasks.

‍
It's not an interface that simulates a conversation, but a system that works together with people to solve real problems.

This service was built for you, if you are

Innovation Leader
Hai il mandato di portare l’AI in azienda. Hai già fatto un primo pilot: in sandbox ha funzionato. Ora però ti chiedono quando andrà davvero in produzione e la risposta non è così semplice.
Il tuo problema
Sai che la tecnologia c’è, ma manca un metodo per renderla davvero affidabile e adottata su larga scala.
accountable
CX/ITSM
Manage customer service or IT support. You've seen what conversational agents promise and you also know what happens when they don't work: wrong answers, blocks, frustrated users.
Your problem
You are not looking for demos, but solutions that really improve the experience, without making it worse.
Digital Leader
Lead the digital transformation. You have clear budgets, teams, and goals. But every AI project takes months before it becomes something really usable, and time is often wasted in unstructured iterations.
Your problem
You need a replicable process, not projects that start from scratch every time.

Our Evidence-Based Method: From Research to Deployment

Let's not jump into the void. 


‍

Our path follows the Double Diamond phases (Discover, Define, Develop, Deliver) adapted to the logic of Artificial Intelligence, reducing risks and maximizing ROI before writing a single line of definitive code.

01.

Discover

‍

Find out the right problem to solve

Through interviews with stakeholders, process observation and data analysis, we map the current state and identify the automation opportunities with the highest impact.

‍

Output:

  • AS-IS Blueprint: map of the current process and of the breaking points;
  • Opportunity Map: areas with greater automation potential.

02.

Define

‍

Design the context in which the agent will operate

During co-design workshops, we define priority processes and future experience. Every agent must know who they are, what they can do, what they cannot do and how to measure success: we define rules, KPIs, the most suitable AI model and the Value Matrix to calculate the ROI. Without this phase, the agent would act in a vacuum.

‍

Output:

  • Agent Definition File: defines the agent's identity, abilities and success criteria;
  • Context Engineering Document: map rules and parameters to make the agent act consistently.

03.

Develop

‍

Build with specialized agents in parallel

We create an interactive prototype in an AI sandbox and quickly test it with real users to validate effectiveness, accuracy and usability, before writing a single line of definitive code. Tasks that would normally take many hours can be completed in a short time by a team of specialized agents.

‍

Output:

  • Working prototype: interactive version of the agent for real tests;
  • MCP Semantic Server: infrastructure that allows the agent to understand and manage the context in an intelligent way.

04.

Deliver

‍

Commissioning and control plan.

We manage the pilot deployment on a controlled group of users, immediately measuring adoption and accuracy. Go-live is not the end, but the beginning: we define performance metrics, operational limits, supervision plan and continuous improvement cycle.

‍

Output:

  • Monitoring Plan: the dashboard that always shows if the agent is working as expected.

Our Success Story

Conflux case study for Italo - UX design booking experience
Case Study
Italo
Conflux for Chiesi
Lorem ipsum color sit amet, consectetur adipiscing elite. Curabitur faucibus, mi sed vehicula ornare, lectus neque commodo massa, eget luctus magna risus nec arcu.
Watch the case study
Conflux case study for Enel Lumiè - UX energy simulator
Case Study
Enel
Conflux for Biogen
Lorem ipsum color sit amet, consectetur adipiscing elite. Curabitur faucibus, mi sed vehicula ornare, lectus neque commodo massa, eget luctus magna risus nec arcu.
Watch the Case Study
Companies that invest in a customer-oriented CX strategy gain a tangible competitive advantage, improving customer loyalty and increasing profitability by 60% compared to competitors that do not follow this approach.
Luigi Greco
Marketing and Sales Manager at Conflux

Our Key Competencies

Benchmark

We compare services with those of competitors and comparators, identifying areas for improvement and opportunities for growth.

Learn more
Service Design

We design services that improve the user experience and optimize our business processes for better results.

Learn more
Prototype Test

We test your prototypes with real users to improve usability, reduce launch risks and create more effective products thanks to our process.

Learn more

What do we deliver
Not what do we promise

Each Agent Service Design project, carried out with our methodology, produces 4 tangible deliverables, regardless of the AI technology chosen.

01.
Agent's Identity Document
Agent Definition File
Who is your agent, what he can do, where he stops, how he behaves in extreme cases. The document that makes the agent explainable to your team, your customers, your CEO.
02.
Context Operational Manual
Context Engineering Document
The instructions that teach the agent to read the situation properly before taking action. The difference between an agent who makes the wrong hires and one that operates with precision.
03.
Smart Connector to Business Systems
MCP Semantic Server
The translator between the language of AI and that of the systems you already use — CRM, ticketing, ERP, database. Without this, the agent lives in isolation.
04.
Control and Supervision Plan
Monitoring Plan
The dashboard that shows if the agent works as designed. Performance metrics, intervention thresholds, escalation process towards the human team.

Agent Design Canvas

A visual and operational tool that transforms abstract concepts into concrete design decisions.

‍

The questions that the Canvas asks you to answer:

01.
Who uses this agent? At what point in your work?
02.
Which problem does it solve and which one should it not touch?
03.
What does success look like? How do we measure it?
04.
What systems does it need to reach? What limits does it have to respect?
05.
What happens when the agent makes a mistake?
Arrow
Download the Agent Design Canvas
Free PDF — no registration required

Not general advice. A method for systems that last.

Conflux doesn't bring AI into the company. Bring agentic systems designed to last — in real work, with your real users, on your real data.

Service Design

Let's start with who will use the system and what problem they really have, not from the list of things that AI knows how to do.

‍

So: let's design processes, not just prompts.

Context Engineering

let's design the agent's informational context: what he knows, what he doesn't know, where he stops, what success looks like.

‍

So: the agent operates precisely, not by trial and error.

Multi-Agent Orchestration

(coordination of specialized agents) — we use teams of agents that work in parallel, with automatic gate qualities and human supervision at critical points.

‍

So: speed and control together, not in conflict.

Nessun
vendor lock-in
You have a mandate to bring AI into the company. You've already done a first pilot: in sandbox it worked. But now they ask you when it will actually go into production and the answer is not that simple.
Guardrail
by design
Every agent has explicit operational limits, interruption mechanisms, escalation to the human team in critical cases. Zero infinite loops.
Evidence-based,
not hype-driven
We don't promise AGI. We promise agents who solve measurable problems in the defined time with you.

A unique approach: Design + Flexible Platform

Our value is not only in the method, but in the technology it enables. We combine the strategic consultancy of Service Design with an open platform, to give you maximum control.

‍

  • Service - The Method: We provide guidance. We apply Service Design, UX Research and co-design to design the agent on the correct use case and define KPIs to measure its success.
  • Platform - The Technology: We provide speed and control. We use whitelabel, SOC2/GDPR-ready and interoperable platforms. There's no lock-in: you choose the best AI models (Bring Your Own API) and keep full control over data, costs and privacy.

KPI & Continuous Governance:

Beyond Go-Live

We are not selling a project, but a platform that improves over time. AI isn't static: we measure agent impact like any other business process.

‍

  • Impact Measurement (KPI): We track clear success metrics, such as the accuracy of the answers, the reduction in the average resolution time (MTTR), the percentage of human escalation, and the cost per task.
  • Continuous Governance: We provide real-time dashboards to monitor SLA, latency, and costs. We manage the tuning of the prompts, the updating of the models and the adaptation of the escalation rules for continuous improvement.

From Blueprint to Operational Agent

The concrete results

Let's bring tangible results to the table. Our journey transforms a process analysis into an operational and measurable co-pilot in a few weeks.

‍

  • AS-IS & TO-BE Blueprint: A clear map of your current processes and how AI will make them more efficient.
  • Interactive Prototype Tested: A sandboxed agent, validated by real users to ensure adoption.
  • Operational Copilot (Pilot): Your tailor-made AI agent, ready to use for a first group of users and integrated into your systems.
  • Governance Dashboard: A control tool to monitor KPIs, costs and agent performance in real time.
Arrow
Il nostro manifesto agentico