Automation & AI

AI doesn't just change your tools. It changes the way your company works.

Find your way in → See the 15 offers

The real question

Not “which AI?”, but “which work do we want to improve, and with what level of control?”

The risk is not running out of ideas, but starting in the wrong place.

1

Buy a tool

A licence defines neither the work to improve nor the rules of use.

2

Launch POCs

Without metrics or an exit decision, the demo stays isolated.

3

Multiply ideas

The portfolio grows before priorities are set.

4

Then look for value

Costs, data, integrations and responsibilities come too late.

Seven dimensions

Processes Value Data Integration Usage Risks Governance

There is no single starting point

We start from what already exists; whatever the starting point, we first look for the next useful decision.

01

A roadmap to define

Priorities, budget, governance, organisation

02

Tools already deployed

Licences in place, usage to optimise

03

POCs already launched

Continue, industrialise or stop

04

A specific process to transform

Simplify, integrate, automate

05

Solutions already in production

Assistants, automations, agents to evolve

06

A portfolio that is spreading thin

Initiatives, suppliers, teams to coordinate

Our method

The right reflex: fix the process before talking about AI

01

Simplify

Remove what is useless, clarify the process.

02

Automate

Structure the data, connect tools and rules.

03

Add AI, if useful

Understand, summarise, handle variability.

The smartest solution is sometimes a simplification, a rule, an API or a deterministic workflow. Saying “you don't need AI here” can be a perfectly useful conclusion.

Start from real work, not technology

Employee Tools Processes Data Information system

The visible experience

  • Experience, productivity, simplicity
  • Collaboration and new ways of working
  • Support and adoption

The foundations required

  • Architecture and integration
  • Governance and security
  • Organisation and operations
  • Data, rights, responsibilities

Unified workplace →    New ways of working →

New capabilities, new responsibilities

What AI can bring

  • Process faster and at scale
  • Understand unstructured data
  • Generate, summarise, assist
  • Assist or act in a variable environment

What must be controlled

  • Quality, errors and evaluation
  • Data, permissions and security
  • Costs, suppliers and maintenance
  • Human controls and responsibilities

The acceptable level of autonomy depends on criticality, measured reliability and reversibility.

Types of solution

Which type of solution?

Automation

Focus: the process: known rules, reproducible result

Examples: routing, reminders, synchronisation, consolidation

AI knowledge assistant

Focus: knowledge: find, summarise, cite sources

Examples: procedures, documentation, contracts

AI work assistant

Focus: the individual

Examples: emails, calendar, meetings, documents, tasks

AI business agent

Focus: a function: goal, variable situations, defined actions

Examples: qualification, coordination, checks, multi-step processing

Automation or agent: the boundary is variability, not technological prestige. An LLM can be a component of several solutions; it does not define the product on its own.

Measure value on a real case, costs included

01

Before

Time, volumes, lead times, baseline quality, current costs.

02

During the test

Representative cases and exceptions, usage and human corrections.

03

To decide

Net gain, full cost, acceptable quality: continue, adjust or stop.

Time freed up only becomes a saving if the organisation can convert it.

The catalogue

4 families, 15 offers

Frame & decide · Build & automate · Adopt & transform · Steer & evolve

01

Frame & decide

“What should we do?”

4 to 8 weeks

AI strategy & roadmap

A prioritised, budgeted, organised and governed Automation & AI roadmap.

“What strategy and what budget to launch AI?”

Offer details +

For whom: Executives, executive committees, CTOs, CIOs, transformation leaders.

What we do: Analyse challenges, populations, tasks and IS; assess foundations and governance; qualify uses, tools and dependencies; prioritise and estimate budgets; define organisation, KPIs and launch.

What you get: A 12 to 24-month strategy and roadmap, indicative budget, organisation, governance and KPIs, launch plan, high-level business cases.

Six decisions covered: Priorities, budget, organisation, foundations, governance, launch.

0.5 to 1 day + preparation/debrief

AI exploration workshop

Quickly surface useful cases without diving into technology too early.

“Everyone is talking about AI, but what could it change for us?”

Offer details +

For whom: Executives, business departments, teams.

What you get: Contextualised opportunities and use cases, value hypotheses, initial prioritisation, recommended next steps.

2 to 4 weeks

AI tool selection

Choose the right AI tools by population and by task, not by trend.

“ChatGPT, Claude, Gemini, Copilot… what to choose, for whom and for what?”

Offer details +

For whom: Executives, CIOs, CTOs, Digital Workplace.

What you get: Comparative tests on real cases, usage-tool-licence matrix, recommendations, usage rules and roadmap.

3 to 6 weeks

Business, automation & AI audit

Identify where the company loses time and which levers deserve to be launched.

“Where are we losing time and where should we really invest?”

Offer details +

For whom: Management, business units, support functions, IT.

What you get: Friction map, prioritised opportunities, operational roadmap, estimated ROI and Quick Win, simplification / automation / AI levers.

2 to 6 weeks

AI control & governance audit

Check that AI uses are under control before extending them.

“Are our AI uses and solutions really under control so we can keep moving forward?”

Offer details +

For whom: CTOs, CIOs, DPOs, CISOs, architecture, AI/Data leads.

What you get: Control diagnosis, risks and gaps, governance principles, remediation plan.

02

Build & automate

“How do we make it happen?”

2 to 4 weeks

AI Quick Win (POC)

Test an idea on a real case and decide whether it deserves investment.

“Does this idea really work for us?”

Offer details +

Condition: Sponsor, metrics and exit decision are mandatory (to avoid “POC theatre”).

What you get: A demonstrable solution, results and limits, estimated or measured gains, a path to production.

3 to 8 weeks

Process automation

Reduce workload, errors and lead times with a reliable, integrated solution.

“Why is someone still doing this task manually?”

Offer details +

For whom: Small and mid-sized companies, operations, support functions, business units.

Principle: AI is not added if a deterministic rule is enough.

Technologies: Power Automate, n8n, Make, API, scripts, RPA, OCR/IDP, LLM if needed.

4 to 8 weeks

AI knowledge assistant

Make internal knowledge usable in natural language, with sources and access rights.

“The information exists, but finding the right answer takes too long.”

Offer details +

What you get: An operational assistant, connected corpus and access rules, test set and quality measurement, sourced answers.

Technologies: LLM, RAG/search, SharePoint, Confluence, Drive…

4 to 8 weeks

AI office assistant

An assistant centred on the person, connected to their daily work.

“I want an assistant that knows my context to help me day to day.”

Offer details +

For whom: Executives, employees, knowledge workers.

Principle: No broad access by default; permissions, approvals and safeguards.

Technologies: ChatGPT, Claude…, M365 / Google Workspace, MCP/API.

6 to 12 weeks

AI business agent

Entrust part of a business process to an agent, with bounded and supervised autonomy.

“Can we give the system a goal and let it handle part of the process?”

Offer details +

For whom: Business units, operations, IT/CTO.

Principle: Progressive, bounded autonomy, human-in-the-loop depending on criticality.

What you get: An operational agent or advanced pilot, explicit level of autonomy, tests, safeguards, supervision, documentation.

03

Adopt & transform

“How do we create real usage?”

2 to 8 weeks

AI assistant rollout

Turn an AI licence into useful, safe and measured usage.

“We have the tool. How do we avoid it being unused or misused?”

Offer details +

Principle: Start from real work, not buttons. Not a generic vendor training.

Technologies: ChatGPT, Claude, Gemini, Copilot, Le Chat…

4 to 12 weeks

AI: from potential to value

Evolve tasks and routines to achieve lasting value.

“We have ChatGPT/Copilot, but the way we work has barely changed.”

Offer details +

Approach: For each task, remove, simplify, automate, augment or keep it human.

6 to 12 weeks

AI coaching for executives

Help an executive form their own conviction by working on their real files.

“I need to learn to work with these tools and sponsor the subject with discernment.”

Offer details +

For whom: Executives of small and mid-sized companies, executive committees.

What you get: Personalised routines and use cases, journal / playbook, organisational recommendations.

04

Steer & evolve

“How do we stay on course?”

Monthly, recurring

Outsourced AI leadership

A senior outsourced Automation & AI function to steer the roadmap.

“Who sets priorities and makes sure we invest in the right places?”

Offer details +

What you get: Periodic committee, roadmap and portfolio, trade-offs (including build vs buy), KPIs and reporting.

Monthly, recurring

AI solutions evolution

Keep improving the automations, assistants and agents already in service.

“Who keeps improving our solutions once they are in place?”

Offer details +

What you get: Periodic review, prioritised backlog, improvements (models, prompts, workflows, integrations), tests, reporting.

Where to start

Three ways in

Explore

When the subject is still unclear

AI exploration workshop

0.5 to 1 day

See the offer →

Decide

When you need to prioritise and govern

AI strategy & roadmap

4 to 8 weeks

See the offer →

Prove

When an idea deserves a real test

AI Quick Win

2 to 4 weeks

See the offer →

“Licences change little about the work” → AI assistant rollout

Not sure which way in?

A first conversation is enough to choose it together.

Book a meeting →

References

What we have delivered

Automation & AI scoping delivered

Large regulated European organisation

6 weeks of immersion
7 stakeholders
30,000+ tickets analysed

6 qualified use cases, 3 Quick Wins, a roadmap of about 9 months.

Audit delivered

Multi-site distribution SME

8+ frictions mapped (double entry, CRM/tool silos)

Quick Win on email sorting and routing; structural projects dependent on APIs and the DMS/CRM.

Scope

Our scope

Ellipsia delivers: transformation, Digital Workspace, processes; automation, AI scoping, strategy and governance; use cases, adoption, integration, assistants and agents; technology choices, supplier management, value measurement; target configuration and support for client administrators.

Ellipsia defines and guides; operations remain the client's responsibility.

Let's talk

Which decision would you like to shed light on first?

01

The choice

What is the first subject that needs to improve?

02

The scope

Which process, team or solution is concerned?

03

The expected proof

What observable result will make you continue, adjust or stop?

Book a meeting →

30 min · Microsoft Teams · No commitment

Book a first conversation

Calendar not showing? Open the booking page

Nous utilisons des cookies pour améliorer votre expérience sur notre site. Rien d'autre.
Accepter
En savoir plus