AI, delivered.Helios · from Ludia Consulting
Before you fund it. A clear plan for AI, and a clear path for every person who works alongside it.
Helios maps the work in one business line and puts three things on every row: the KPI it moves, its build and run cost on Microsoft, and the person who owns it. Your leaders then fund what pays back, measure first where there’s no baseline, and take the cheaper route where a workflow or a person does the job.
1How it works
Helios doesn’t start from scratch. We come to the first meeting with your map already pre-filled. Then your people tell us how the work runs, your executives make the calls, and our architects and Ludia’s proprietary agents do the drafting. You leave with what our customers value most: a clear direction for your AI strategy, and for its execution.
Watch it in under two minutes
Why most AI is hard to put to work, how Listen, Strategize and Execute fix it, and what you leave with. Captions are on, so it plays with the sound off.
We arrive with your industry’s map pre-filled, then listen. Your leaders and the people who do the work tell us how it really runs, and Ludia’s proprietary agents turn every conversation into a draft map of your business line.
We sit with your leaders and the people who do the work, about an hour each. Their words go on the map.
Our consultants run every interview through Ludia’s proprietary agents to draft the map: each function, the work inside it, and where its data lives.
The map of one business line, in your people’s own words, and where its data lives.
Your experts and our solution architects decide what each piece of work should become, which KPI it answers to, and in what order it happens.
Our solution blueprint and solution architecture teams sort every task with your experts: an AI agent, a rule-based workflow, or a person. Each call is checked against Microsoft’s Success by Design and business process libraries.
Your executives set the KPI each business line is held to, and we write down today’s baseline.
Every row gets a named owner and a wave: start now, next, or later.
A roadmap your leaders own: agent, workflow or person on every row, the KPI it answers to, a named owner and a wave.
Every investment is tied to a number before it’s funded, and measured after it’s built. Cost and result sit on the same row, so nobody has to take the return on faith.
Before anything is funded, every agent gets a full estimate:
KPI, cost and outcome on one row, so your leaders can fund it, measure it first or take a cheaper route.
A clear direction for AI, and the proof it’s working. What to build, in what order, what it costs, which KPI it moves, how each person’s work gets better, and the numbers that show it paid back.
Ludia’s proprietary method. Our process, our agents and our tools, built on Microsoft’s Success by Design and business process libraries and on hundreds of years of Ludia know-how.
Three stages, step by step
Leaders first, then the people who do the work, about an hour each. Every function goes through the same structured conversation, so nothing depends on who asked. We bring your map pre-filled for your industry and your company, and the interviews correct it.
Ludia’s proprietary agents, the ones we built to run Helios on our own business, turn each interview into a draft map. They pull out the functions, the work inside each one, what it produces, and the system its data lives in. A Ludia consultant checks every row before your team sees it.
A process stays with a person (safety calls, regulatory sign-off, pricing), and a workflow follows rules, so it needs no AI model. An agent handles judgment, language or synthesis, and a named person owns what it produces. Our solution blueprint and solution architecture teams make each call with your subject-matter experts, checked against Microsoft’s Success by Design and business process libraries.
Your executives set the KPI each business line is held to: availability, quote turnaround, utilization, whatever the line already reports. We write down today’s baseline and its source; each function’s measure rolls up to it and each agent inherits it. The whole Decide phase is built around this, because a KPI your own leaders set survives the first quarter.
Every row gets a value and ease score on Ludia’s written rubric, a named owner and a wave: now, next or later. A row that’s waiting on a data fix or an integration says exactly what it’s waiting for.
Before anything is funded, every agent gets a full estimate, worked out from who uses it, how often, and what it reads. First we check the rows against the agents Microsoft already ships in Dynamics 365 and Microsoft Copilot, so you don’t build what you already license.
KPI, cost and outcome on one row. Your leaders fund what’s worth more than it costs, measure first where there’s no baseline yet, and take the cheaper route where a workflow or a person does the job. The readout is a working session with the people who own the numbers.
What you keep
They’re working files your team keeps editing, and they stay yours whoever builds the first wave.
The radial view of the business line, sized by where the work is. The picture your executives point at, and the one that survives being forwarded.
One line per row: mechanism, who it serves, system of record, value and ease, wave, owner and estimate. Each row also carries a draft brief, so it’s halfway to being built.
From the work to the function to the business line KPI, with today’s baseline and its source. Your leaders set it in KPI alignment.
Rows in waves with target quarters, and the KPI, cost and call for each. The first wave leaves as scoped work with named owners.
What Helios is not
Helios produces the map and the plan. The build follows, separately. Anyone promising deployed agents from an assessment is selling you a demo.
We don’t score your data maturity. We show which rows are blocked on data, which turns out to be more useful and a lot cheaper.
Copilot itself mostly comes in the box. Helios is about the thirty to sixty agents and workflows that don’t.
Workshops produce lists. Helios produces a scored, owned, sequenced plan with a KPI and a cost on every row.
Stage 3, after the readout
Ludia Consulting builds the first wave on Microsoft Dynamics 365 and the Microsoft agent platform, using Ludia’s AI Implementation Methodology. Rows move from proposed to in build to live in the same workbook, and Copilot Studio reports the credits each agent uses, so the second wave is sized from your own numbers.
System of record exists and the process is stable. Days to a few weeks each.
Each live row reported against the KPI your leaders set, next to what it cost.
The next business line is added to the same map. Nothing is rebuilt.
2Industries
Five worked maps, one per industry. Each names the business lines we expect to find and the numbers leaders there are already held to. The interviews then correct it, in your people’s own words.
Leaders here are held to
Plant operations · Maintenance and outages · Compliance and safety · Commercial and owner
Open the energy map → Discrete, process, contractLeaders here are held to
Order to quote · Plan and produce · Quality and compliance · Aftermarket
Open the manufacturing map → Consulting, engineering, ITLeaders here are held to
Pursuit · Delivery · Practice operations · Client success
Open the services map → Dealers, rental, product supportLeaders here are held to
Machine sales · Rental fleet · Parts · Service
Open the equipment map → State and local governmentLeaders here are held to
Constituent services · Infrastructure and assets · Administration and finance · Workforce and governance
Open the public sector map →A different industry? The method holds and the business lines change. Tell us how your business is organized →
The four business lines, here
The map
This map. rows across functions, serving distinct roles: % agent, % workflow, % process. have nothing to read from today, which decides how much can start in the first wave.
What stays human here
What doesn’t change
Listen, strategize, execute: interviews, analysis, agentic distribution, KPI alignment, roadmap, estimation, and the business case. What each business line is called is one of the answers, never an input. How it works →
In roughly a 60 / 30 / 10 shape. What lands in each bucket changes completely between these five maps. The shape barely moves.
A function that produces something countable gets measured on cycle time, rework and volume per person, whether it’s a shift turnover, an 8D report or a warranty claim. Each agent’s cost sits next to it. KPIs and ROI →
A risk or compliance function asked to review an AI portfolio looks first for evidence that somebody thought about where the model should not be. A map that names those rows, with the role that owns each, gets through review. A list of agents comes back six weeks later with a request for exactly this analysis.
Agent 365 as a control plane with a central agent registry, Entra Agent ID giving each agent an identity, and Purview and Defender extended across agents. The governance question has a product answer now, which is newer than most risk teams realize. For IT leaders →
3KPIs and ROI
An agent takes on the KPI of the business line it serves. If that line commits to a lift your CFO can put a value on, that value is what the agent’s build and run cost has to beat. Both numbers sit on the same row, so the return is arithmetic, not faith.
Set by your executives for each business line, with today’s baseline and where that number comes from.
Estimated per agent from what the interviews recorded: who uses it, how often, and what it reads.
Every row lands in one of three places, with the reasoning written next to it.
It goes into the first wave with a named owner and a target quarter.
The first wave starts counting, so the second is sized from your numbers instead of ours.
A workflow carries no AI model cost. Some work should stay with a person, and the map says which.
Across the maps we’ve built, our own included, roughly four rows in ten turn out to be workflows or work that stays human.
That’s money you don’t spend on AI, and it’s why the other six are easier to believe.
From the work to the KPI
Every function on the map names the things it produces: bids, work packages, cases, field reports. Those can be counted on cycle time, rework and volume per person, and they roll up to the KPI the business line is already held to. Because the map is built function by function, the measurement comes out of the map instead of being bolted on afterwards.
Hours or days from the trigger to the finished piece of work. The one your people already feel.
How much each person produces in a period. Capacity, without touching headcount.
How many passes before the work is accepted. Often the bigger win, and the one leaders underestimate.
The share done right the first time, or on schedule. The risk and commitment story.
The rule that keeps it honest: every row’s measure rolls up to a function measure, and every function measure connects to a business line KPI you already track. A row that connects to nothing you measure is usually a low-value row, and it gets cut.
The arithmetic
Your numbers go in. We supply the method and show every assumption, so the argument in the room is about the inputs, which is where it belongs.
The volume the KPI covers, times the improvement your leaders commit to, times what one unit of that improvement is worth to the business.
Your executives, in KPI alignment, from today’s baseline.
Build effort, spread over the period you choose, plus the monthly cost to maintain and run it on Microsoft, plus any new seats the row needs.
Ludia’s estimate, from what the interviews recorded.
Worth more than it costs: fund it. No baseline yet: measure first. A workflow or a person does it for less: take that route.
Your leaders, at the readout, row by row.
The estimate, per agent
Microsoft bills agents on consumption, in Copilot Credits, and says the amount depends on how an agent is designed, how often it runs and what it uses. A Helios row already records all three, which is what makes a per-agent estimate possible where a per-tenant one is a guess.
Runs per month from the trigger and the people served; the size of each run from the inputs it reads and the output it writes.
Compare agents on the same footing before any of them is built.
Usage priced the way Copilot Studio bills it, as a monthly range, and cross-checked against Microsoft’s agent usage estimator, plus what it takes to maintain the agent as your data and processes change. Beside it, seats: which roles need a Copilot or E7 seat, and which don’t.
Put a number next to the KPI it has to beat.
Days, weeks or months, set by the mechanism and by whether the system of record exists today.
Size the first wave against the capacity you have.
A wave (now, next or later) and a target quarter, sequenced on named dependencies: a data fix, an integration, a record that doesn’t exist yet.
Plan the whole year.
System of record exists, process is stable. Days to a few weeks each. Startable the week after the readout.
Waiting on one named thing: a data fix, an integration, a process that has to settle. Weeks to a couple of months each.
Real work with no deadline yet. It stays on the roadmap so nobody rediscovers it next year, and it’s re-estimated when its dependency clears.
It firms up when the row goes into build. Copilot Studio then reports the credits each agent uses, the workbook keeps that next to the estimate, and the second wave is sized from evidence the first wave produced. A fixed price per agent, quoted before anyone has touched your data, is a guess with more decimal places.
Microsoft’s agent usage estimator ↗ How Copilot Credits work →After the first wave
By business line and by agent, workflow or person, against the waves your leaders agreed.
Each function’s measure against the baseline written down in KPI alignment.
Consumption and seats against the forecast, per agent.
Proof your leaders prune the plan as well as grow it, plus what moved from next to now and what’s blocked, and why.
What we won’t put in a business case
CFOs are trained to spot a made-up number, and one is enough to sink a whole plan. So we hold three lines.
Baseline, volume and improvement band, each with its source named: your own figure, a range we’ve seen on comparable work, or an assumption to test in the first wave.
What the map found is countable. What building it might return is modeled, and we label it that way every time.
Where nobody measures something today, we say so and make measuring it part of the first wave.
4Microsoft engine
Helios decides what each piece of work should become. Microsoft’s tools run it. Because the map covers the whole business line, it uses them together: you see the complete picture before you buy anything, and the build becomes one engine, governed end to end.
Most agents on the map land here, so nobody has to learn a new tool.
Agentic distribution decides which tool each row needs.
Every row names its system of record, so each agent reads the right source.
Every agent gets an identity and an owner before it’s built.
One cloud, one identity, one place to govern it.
Names and statuses as of October 2026. Product icons are Microsoft’s official icons, shown unmodified; Microsoft has none yet for Entra Agent ID or Purview.
Ludia is a Microsoft Solutions Partner for AI Business Solutions and a FastTrack Portfolio Partner. Our founder and CEO played a key role in shaping Dynamics 365. So the map ends in a license line, a hosting decision and a backlog your team can own.
Where the data lives
On the industry map in section 2, of rows have no system of record today: a cost model in a folder of spreadsheets, an outage scope list emailed round with a version number in the file name. Those need a platform decision before they need a model, and naming them is how a map turns into a platform roadmap.
Rows that read from a system outside Microsoft, such as a CMMS, an ERP or a historian, reach it through a connector or an API, and that integration effort is part of the row’s build estimate.
Where these numbers come from. Whichever illustrative map is open in section 2. Your map counts your own rows.
Build surfaces
Microsoft publishes the ingredients. The decision tree below is ours, written in Microsoft’s vocabulary, and it’s why a Helios row can be costed before it’s built.
Safety-critical judgment, regulatory sign-off, pricing calls, second-person verification.
Cost: nothing. It’s also what makes the rest of the portfolio defensible. A plan with stated exclusions passes a risk review that a bigger plan without them doesn’t.
Deterministic steps, no judgment. Microsoft’s words: “Agent flows are deterministic. They execute actions or tasks following a rule-based path. The same input always produces the same output.”
In Helios. Billed per action, with no AI model unless you add an AI step. Agent flows bill through Copilot Studio capacity; cloud flows need a Power Automate license. They’re cheap, fast and easy to trust, so we sequence them first on purpose.
Agent flows overview ↗Declarative agents use the same orchestrator and models as Microsoft 365 Copilot and inherit its compliance posture. Custom engine agents bring their own orchestration and model choice. They need their own hosting unless they’re built in Copilot Studio, which hosts them in Microsoft 365.
In Helios, this is a scoring input. A declarative agent can’t start itself; it only runs when a person asks. So any row that has to fire on an event is a custom engine agent, which changes its cost, its hosting and the compliance conversation. We decide that on the map.
Agents for Microsoft 365 Copilot ↗ Microsoft Foundry Agent Service ↗“Route deterministic tasks to rule-based logic instead of premium language models to maximize efficiency.”
Microsoft’s guidance makes the same argument the workflow column does. The expensive mistake is turning everything into an agent, and a plan that starts from the technology makes it by default.
Integrate, manage and operate AI agents ↗Secure by construction
Your risk committee will read this section first. Microsoft now has a product answer to agent governance, and Helios hands your risk team the evidence in a format they already use.
Microsoft Entra Agent ID makes each agent a governable principal, and it’s available to every Entra customer. Conditional access, lifecycle management and access reviews apply to agents the way they already apply to people. An agent that can’t be revoked doesn’t get deployed.
Agent 365, which Microsoft calls the control plane for AI agents, keeps one inventory of every agent in the tenant, including ones built outside Microsoft, with telemetry and alerts across the fleet. Extending Entra security to agents needs Agent 365, which Microsoft 365 E7 includes.
Purview for data loss prevention and data security posture, Defender for threat detection, Intune for device posture, extended across agents rather than replaced. There is no parallel security stack to staff.
Copilot connectors index external content with its access control lists, so Copilot only shows items to people who already have access. An agent running as the user sees what that user can see. An agent running on a maker’s credentials or its own identity sees what that identity is granted. So “what does this agent read, and as whom” is a governance decision on the map, not a build detail.
Copilot connectors ↗That phrase is Microsoft’s, from its agentic maturity model, and it’s the pattern Helios recommends on each agent: the agent drafts, and a named person reviews and owns the outcome. The Dynamics 365 Sales Qualification Agent ships with the line “the agent doesn’t replace your judgment or decision-making process”. Business Central’s Payables Agent introduces an agent supervisor, which makes human-in-the-loop a job title.
What stays human on a Helios map →The license
Microsoft 365 E7, which Microsoft calls the Frontier Suite, has been generally available since 1 May 2026. In Microsoft’s words, it “unifies Microsoft 365 E5, Microsoft 365 Copilot and Agent 365 into a single solution powered by Work IQ.” Prices vary by agreement, so we leave them to your Microsoft account team.
Included in E7 · formerly Microsoft 365 Copilot
The assistant inside the apps your people already use. Mostly comes in the box.
In Helios. Almost never a row on the map. Turning Copilot on isn’t a portfolio; Helios is about the thirty to sixty agents that don’t come in the box.
Microsoft 365 E7 ↗
Included in E7 · GA 1 May 2026
A central registry of every agent in the tenant, Microsoft and non-Microsoft, with access control, observability and security over the fleet. Licensed per user, not per agent. Registry sync with Amazon Bedrock and Google Cloud is in preview.
In Helios. The answer to the question every executive asks before funding: what happens when there are forty of them? Because the map counts agents before anything is built, the Agent 365 conversation happens while the number is still a forecast.
Agent 365 overview ↗Gives an agent an identity that can be authenticated, authorized, governed and revoked like a person’s. Agent identity blueprints let one policy govern a family of agents, and an agent’s user account lets it hold a mailbox and a calendar where the work needs one.
In Helios. Each agent on the map carries the roles it serves and the system it reads from, which is exactly what an identity and access review needs. Entra Agent ID gives it an identity; Agent 365 governs the fleet. Two products, two jobs, and we keep them separate.
What is Entra Agent ID ↗
Not a SKU · APIs GA 16 June 2026
The workplace intelligence layer agents reason over, built across Microsoft 365 and external systems with permission-aware governance. Four surfaces: Chat, Context, Tools and Workspaces. Every tool call is logged.
In Helios. It’s what makes a Microsoft 365-grounded agent viable without a bespoke retrieval pipeline. Work IQ inside Copilot Studio is still in preview, and we say so on the row rather than in a footnote.
Work IQ overview ↗Agentic work bills on top of the seat, in Copilot Credits. Any total-cost story that leaves consumption out won’t survive a procurement review, which is why Helios estimates it per agent.
E7 is a license. The Microsoft Frontier program is early access to pre-release AI features: no SKU, though business users need a Microsoft 365 Copilot license and some features carry usage-based costs. If a vendor says they’ll “deploy Frontier”, ask which one they mean.
The surface your people touch
Naming the surface a row lands on is part of scoring it, because it decides the license, the governance model and whether anyone has to learn something new.
Microsoft CopilotFormerly Microsoft 365 Copilot. AI inside Outlook, Teams, Word
and Excel, grounded in your own work.
Copilot ChatSecure AI chat for everyone at work. Where a simple agent shows up
with nothing new to learn.
Copilot CoworkTakes on multi-step work across mail, meetings and documents, and
asks before each action. GA since June 2026, billed by use in Copilot Credits.
Copilot SearchSearch across Microsoft 365 and connected systems. Often the honest
answer for a row somebody assumed needed an agent.
Copilot NotebooksGrounded working sets over your own content, for rows whose
system of record is still documents.
Researcher and AnalystCited research from the web and your work, and analysis that
runs on your own data.
Copilot StudioWhere rows that need orchestration, custom tools or an event
trigger get built and governed.Product names and icons are Microsoft’s. The icons are Microsoft’s official files, shown unmodified to identify each product.
What it consumes
Each agent on the map carries its trigger, its inputs, its outputs and the number of people it serves. Frequency times users gives a first estimate you can defend line by line, instead of a tenant-level number nobody can source.
Microsoft’s “common currency for eligible Microsoft services with usage-based billing”. Copilot Studio bills credits two ways. Agents built on Microsoft’s “GitHub Copilot harness” are charged for tokens, tools and runtime from the first build; standard agents are charged per event, such as an answer, an action or a grounding call. Bought as packs, pay-as-you-go, or a one-year Copilot Credit Pre-Purchase Plan. Copilot Studio reports the credits each agent uses.
How Copilot Studio bills credits ↗ Agent usage estimator ↗ACUs are bought in the one-year Microsoft Agent Prepurchase Plan. Each pays a dollar of retail usage across Copilot Credit services, Microsoft Foundry, Fabric and GitHub, and purchases are final. They aren’t Copilot Credits, and they aren’t the separate credit pre-purchase plan either. A licensing specialist will spot a document that mixes them up.
Agent Prepurchase Plan ↗Seats work the same way. The distinct roles across the agents tell you who plausibly needs a Copilot or E7 seat, and the workflows and processes say who doesn’t. That denominator is what makes the number hold up in procurement. How the estimate is built →
Dynamics 365
Before we cost a single build, we check every row against Microsoft’s first-party Dynamics 365 agents. Statuses below are Microsoft’s own as of October 2026. Business Applications uses three tiers: preview, production-ready preview, and GA.
Sales Qualification, Sales Opportunity, Sales Research, Recommended Actions, plus the Sales agent in Microsoft 365 Copilot. GA.
Case Management, Customer Intent, Customer Knowledge Management, Quality Evaluation and more. GA, with some features still in preview.
Time Entry, Expense and Approvals agents. Production-ready preview.
Scheduling Operations Agent. Preview.
Supplier Communications Agent, part of the Procurement Agent: production-ready preview. Impact analysis: preview.
Account Reconciliation Agent. Production-ready preview.
Sales Order Agent and Payables Agent. GA, with some capabilities still rolling out.
The older static Dynamics 365 ERP MCP server retired on 1 October 2026. The current one is GA and scopes context to the calling user’s role.
Microsoft’s own guidance is close to the Helios thesis: “What matters most is the foundation: structured data, clear governance, and business logic that allows agents to operate effectively.” That foundation is what the map inventories.
Reference
This stack changes every quarter. Check these pages before quoting a SKU, a status or a date, including anything on this one. Last checked 3 October 2026.
5What each leader gets
The same workbook answers the board, the budget review and the architecture review.
6We ran it on ourselves first
We don’t name clients, but the numbers are real. That engagement maps one business line at a time and is extending.
7Our method
Helios is Ludia’s own method. We created the sequence, the scoring rubric, the workbook and the agents that build it, on top of Microsoft’s Success by Design and business process libraries. Below is what it guarantees. The working detail stays inside the engagement, where your team sees all of it.
What every Helios map does, whoever runs it
Every row starts from a business line and the work people do there. The tool is the last decision, never the first.
Agent, workflow or person on every row, with the reason written beside it. A plan that’s all agents is a sales document.
A written rubric, applied by Ludia and confirmed by your leaders. Two people scoring the same row land in the same place.
Every first-wave row has a named owner who agreed to own the number. A team name doesn’t count.
Each row rolls up to a KPI your business already reports, and each agent carries its build and run cost.
Every row names the system its data lives in, so nothing gets designed around data that isn’t there yet.
Inside the engagement
The parts that make Helios repeatable are Ludia’s intellectual property. Your team works with all of them during the engagement, and keeps everything it produces.
The business lines, functions and systems we expect to find, pre-filled before the first meeting.
Built by function, so every conversation produces the same kind of answer.
They draft the map from the interviews, and a Ludia consultant checks every row.
Value, ease and wave, defined before anyone scores a row.
What each agent needs, what it costs to build, maintain and run, its priority and its impact, checked against Microsoft’s own usage estimator.
One row per opportunity, and the picture your executives point at.
If your architecture board, risk function or procurement team needs to see the method before the first interview, we’ll take them through it in a working session. Talk to Ludia →
8One Ludia
Ludia Labs designed the method and runs Ludia on it. Ludia Consulting runs your engagement and, if you choose, builds what the map says is worth building. Same people, same standards.
Our research and product arm. Ludia runs itself on Helios, so every part of the method is tested on Ludia before it reaches you.
Helios on ludialabs.com ↗One business line at a time, with a business case on every row. You keep the map whether or not you build with us.
This siteA software solutions architecture firm focused on Service Operations, on Microsoft Dynamics 365. The first wave becomes the build, with the people who mapped it.
Helios on ludiaconsulting.com ↗9Start with one business line
Tell us which business line that is. We’ll bring a draft map for your industry, walk you through the three stages, and show you what the business case on each row will look like.
