Artificial intelligence
Artificial intelligence where it makes sense, on processes that work.
If the company's processes already work well, we automate them as they are. If needed, we redesign them with the company or design them from scratch, and on top of them we add the layer of automation and artificial intelligence. Either way, only in the tasks where it saves time or brings a decision forward. It is a service that can be contracted on its own or as part of the project management model.
Our view
The value lies in the process and in the people. The technology is the small part.
On a renewable project, time is lost gathering data that already exists, in reports that arrive late and in packages completed the night before. That is where we put artificial intelligence.
- 01
Process first
An agent on a poorly defined process only speeds up the disorder. That is why we review the process before automating it and, if needed, put it in order first.
- 02
Where decisions arrive late
We start where a decision is taken with outdated data: a cost deviation, a milestone that slips, a permit about to expire.
- 03
With the team
Much of the value comes from changing how people work, and we put as much effort into that as into the technology.
What we build
What we deliver is a working system.
- 01
Process agents
They take on the repetitive part of a process and prepare the decision, with tasks such as explaining a deviation, proposing its causes, comparing scenarios or drafting the report.
- 02
Cycle automation
We automate the monthly close, reconciliation with accounting, alerts and scheduled reports, and every change is recorded.
- 03
Control applications
We build portfolio dashboards and scenario simulators, accessed with the company's corporate login.
- 04
Data integration
We bring together the cost, schedule and progress data that today live in separate systems, so that every process works with the same figure and every finished project improves the estimates for the next one.
Project data
Every finished project improves the estimate for the next one.
A company that has built ten plants knows more than its estimates show. Deviations, actual durations and the risks that did occur stay in close-out reports that nobody opens again. We integrate that data so that process agents start from the company's own experience.
01CAPEX deviations by cost item
CAPEX estimate and contingency
Actual overruns by cost item, technology and contract type adjust the estimate of the next project and the size of its contingency.
02Actual schedule durations
Planning
How long foundations, erection or grid connection really took becomes the basis for new schedules and flags optimistic durations.
03Risks that did occur
Risk register
Each materialised risk, with its cost and delay, enters the initial register of the next project and adjusts its probability.
04Actual permitting times
Permitting calendar
How long each authority took for each type of permit adjusts the permitting calendar and the margin against the administrative milestones.
The agent proposes the figure together with the projects it rests on, and the team decides. The more projects go through the system, the sharper the estimate.
Use cases
What it can take on in each area of a renewable project.
These are examples of what an agent takes on in each area: it reads the systems the company already has, does the repetitive part and leaves a draft with its evidence. Anything it writes to a system waits for a person’s approval, and every step is logged.
CAPEX controlDemo
It reads purchase orders, invoices and progress certificates from the ERP and cross-checks them against progress. It reworks the final cost forecast and drafts the note on each deviation, with its cost line and its cause. The controller validates it before it reaches the committee.
Planning
It compares each schedule update with the baseline. It flags the milestones that move, the float that is running out and the dependencies that drag others along. The planner decides what gets replanned.
Risk management
It reads minutes, reports and project emails and proposes new risks, with the sentence they come from. It reminds each owner of the mitigation that is falling due. No risk enters the register until someone accepts it.
Readiness
Before each stage gate, it checks the file against the list: which evidence is there, which is missing and which has expired. It gives the committee the gaps, with who has to close them.
Document management
It extracts dates, conditions and obligations from contracts and permits, and controls versions and approvals. It answers questions citing the document and the page. If it cannot find the answer, it says so.
Governance
It tells anyone on the team, task by task, who carries it out, who approves, who is consulted and who is informed under the RACI. It prepares the agenda for each committee and follows each action until it is closed.
Feasibility and gates
It gathers constraints, assumptions and reference costs into a first draft of the study. At each gate, it checks the criteria against the project’s evidence. The team reviews and signs.
Handover
It assembles the handover file from the list of deliverables and marks what arrives incomplete or unsigned. Whoever receives it knows what they are accepting and with which open items.
Lessons learned
When a project starts, it searches the register for lessons from similar projects and takes them into its plan and its risk register. At close-out, it proposes new ones from the deviations.
Monthly report
It assembles the monthly portfolio report with the closing data, in the format management already uses, and drafts the comment on each variance. The director reviews it before it is sent.
An example
Cost control on a project, before and after.
We have chosen cost control because it is the process where the problem repeats most. The same pattern applies to other processes, such as planning, risks, permits or the monthly report.
See the demo Demo · fictitious data
DiagramFrom deviation to decision
With the agent, the deviation reaches the decision-maker while it can still be fixed.
Today
- The deviation happens
- Month-end close
- Report
- Committee
With the process redesigned and an agent
- The deviation happens
- Alert, with cause and effect
- Decision
- Time to correct
- Who gathers the data
- TodayEach owner sends a figure in a different format. Someone consolidates them by hand.
- With the process redesigned and an agentThe agent reads commitments, invoices and progress from the systems already in place.
- When a deviation is seen
- TodayAt month-end close, when the invoice arrives.
- With the process redesigned and an agentWhen the commitment is signed or the forecast changes.
- What happens next
- TodayIt is explained at the next committee.
- With the process redesigned and an agentIf it is small, the agent redoes the forecast and records it. If it is large, it alerts the Project Manager with the cause, the effect on returns and a plan.
- Who decides
- TodayThe committee, with last month’s information.
- With the process redesigned and an agentThe same committee, with the day’s information.
Rollout
We start with a first project and then move on to the whole portfolio.
- 01
One project
We take one process, on one project and over one full cycle, with success criteria that we agree with the company before starting.
- 02
The portfolio
We extend the system to every project and hand it over to the company's teams with its documentation.
- 03
Operations
The company runs the system. If it asks, we stay on for support and for its evolution as the portfolio grows.
What is built belongs to the company from the first stage.
Before anything is built, the free diagnostic indicates which process is worth automating first. See the diagnostic
Control
Secure and under the company's control.
- 01
Data stays where the company decides
We work with the artificial intelligence models the company already has under contract, inside its corporate tenant and with its identity and access management. If it does not have them yet, we propose the architecture that fits its data policy: private cloud, on-premise deployment or API access with zero data retention. Data is encrypted in transit and at rest, and the enterprise agreement prevents the provider from using it to train its models.
- 02
A person always decides
What the agent proposes stays pending the approval of whoever is accountable under the RACI. Every action goes into an audit log with who approved it, when and with what data, and the agent's permissions never exceed those of the user working with it.
- 03
IT reviews it and takes it over
We hand over the code, the architecture documentation, the operations manual and the automated tests, and deployment follows the company's change management process. The IT team reviews it, maintains it and can audit it.
- 04
No dependence on one provider
An abstraction layer separates the agents from the model. The model is chosen per task by cost, accuracy and confidentiality, and the provider can be changed without rebuilding the system.
Limits
What artificial intelligence does not do.
- It does not decide for the company. It proposes alternatives and prepares the decision, which is taken and signed by whoever is accountable for it.
- It does not fix a badly designed process, which is why the process comes first.
- It does not assume what is not in the company’s data. When a piece of data is missing or unreliable, it flags it rather than estimating it.
- It does not replace the Project Manager or management. It gives them an up-to-date view so they decide earlier and on data, not on outdated reports.
Demo
CAPEX control connected to the financial model.
The demo shows the approved, actual and forecast cost of a fictitious project. Each deviation recalculates returns and reaches management with its cause.
Vari FieldIn development
What happens on site, on the decision-maker's desk the same day.
On a wind or solar farm, schedule and cost are won or lost on site. What is seen there reaches the office days later, in a report someone has to type up. Vari Field is the line in which we take artificial intelligence to the site, so that what the team sees and says becomes project information at once.
- 01
Capture
The site team records what it sees by voice, photo, video or smart glasses, hands free and without forms.
- 02
Interpret
Artificial intelligence sorts each observation as progress, issue, quality or safety, and places it in its schedule activity and its cost item.
- 03
Verify
Each piece of data carries its origin. The site manager confirms what was inferred before it counts in the project.
- 04
Connect
What is confirmed reaches planning, risks, CAPEX control and the management report, without waiting for the weekly report.
Each piece of data carries its origin
- Observed
- Inferred by AI
- To be verified
Designed for the daily work of a renewable site.
Actual progress
Foundations, turbine erection, tracker piling and assembly, trenches and cable laying, measured from the site.
Issues and non-conformities
With photo, location and owner from the moment they are found.
Safety
Safety observations reach whoever has to act on the same day.
Punch list
The list of open items for handover is built during construction and not in the last weeks.
Weather windows
We are studying how to cross the forecast with weather-sensitive activities, such as lifts, concrete pours or work at height, to propose work windows. The decision stays with the site manager.
Vari Field is in development and is not yet a product.
Technology should disappear into the way people work.
We believe that good digitalisation removes manual work and improves the quality of decisions, and that the kind that adds complexity is not worth having.
We love talking about projects.
The first conversation carries no commitment.