Delivered automation projects

How the process worked before

Every month staff had to update contract data against the services actually delivered, change dates and periods, fix document formatting, save the files as PDF and hand them to the finance department.

With ten such contracts the process looks simple. Once there are several hundred, even trivial repeated steps such as changing a value or saving a file as PDF add up to dozens of working hours a week.

What we built

We automated the process from the initial data preparation to the finished document being handed to a person for review. Using the actual service data, the system updates the required information, changes periods and dates, produces the documents from the approved template and prepares the PDF files.

What remains for the employee is to review the prepared documents, check the exceptions and approve them.

What changed

A process that used to take about 50 working hours a week is now prepared and reviewed in roughly 2 hours.

Automation did not remove human control. It removed most of the monotonous steps, so an employee’s time goes into checking documents and making decisions rather than into hundreds of identical clicks.

How the process worked before

Delivery of the digital product depended on someone on the team acting. After payment, the order had to be checked, the right code found and sent to the customer. Delivery could therefore be late outside working hours, at weekends, or whenever the team was busy with something else.

What we built

We built a process that, once a payment succeeds, assigns an available code to the customer automatically and sends it by email. It also monitors the remaining stock of codes, records the delivery status and logs clearly when and to whom each code was issued.

What changed

The customer receives the product immediately after payment and the team no longer checks every order by hand. The process is more reliable, easier to trace and no longer depends on how fast someone responds.

How the process worked before

Before any outreach, the sales team first had to find suitable companies. That meant checking the sector, the city, the website and the contact details, then deciding whether the company fitted the target customer profile at all.

At just 2 or 3 minutes per company, building a list of 500 prospects can take 20 working hours or more.

What we built

We built a process that collects prospects for a chosen market, city, sector and other criteria, and prepares them for the next stage of sales work.

The list is enriched with the company name, a description of what it does, the website, contact details and other information needed for qualification. Defined rules can reject unsuitable companies immediately or sort the contacts by priority.

What changed

Instead of several working days of manual research, the team gets 500 prepared companies in about 30 minutes and can spend its time on outreach rather than data gathering.

A manager can also see clearly which market was covered, on what criteria the companies were selected and how many prospects are ready for the next sales stage.

How the process worked before

Each new customer got a proposal prepared from scratch. Someone copied an earlier document, changed the customer details, services, prices, timelines and other information, then checked the formatting again at the end.

At about 25 minutes per proposal, 20 proposals a week add up to more than 8 hours of repetitive work, along with the risk of leaving in a previous customer’s details, a wrong price or a clause that no longer applies.

What we built

We built an internal proposal process. The employee selects the customer, the services, the pricing and the terms required, and the system produces the document from the approved template.

Customer details are filled in automatically, the right sections are selected, prices are calculated and a consistently formatted document is produced. What remains is to check the content and adjust any individual terms.

What changed

Preparing 20 proposals a week went from about 8 hours to 1.5 hours.

The sales team spends less time copying and formatting and more on the proposal itself, the pricing and the conversation with the customer. Every document also follows the same approved structure.

How the process worked before

Every week or month someone collected the same data again from several sources: the CRM, Excel files, the project management system or other internal tools.

The data had to be copied, merged, checked for missing values and recalculated before the report could be prepared for management. At about 1.5 hours per report, four monthly reports come to roughly 6 hours of repetitive work.

What we built

We built a process that pulls the required data from the selected systems on a schedule, merges it, checks it and prepares it in a consistent report structure.

Where data is missing or a discrepancy appears, the case is flagged for human review. The finished report can be saved as PDF or Excel, sent by email or published in an internal system.

What changed

Monthly report preparation went from about 6 hours to 30 minutes, spent on reviewing results and handling exceptions.

Management receives consistently structured information sooner, and nobody repeats the same collection and copying work each cycle.

How the process worked before

Stock levels were checked periodically, which meant going through hundreds of items looking for the goods or materials approaching a critical threshold.

Checking 300 items twice a week amounts to roughly 12 hours a month on that review alone. The bigger problem appears when a shortage is noticed only a day or two before stock runs out.

What we built

We built a process that monitors inventory data continuously and applies a defined minimum threshold to each item.

The system can weigh not only the current quantity but also average consumption and lead time. If five days of material remain and the supplier needs seven, the responsible employee is alerted before the minimum level is even reached.

The alert states the item, the current quantity and the information needed to act.

What changed

Instead of about 12 hours a month checking the whole warehouse, staff review only the items that need attention.

The main result is not only the time saved. The company learns about a possible shortage earlier and has time to restock before it affects production, service or other work.

How the process worked before

About 60,000 records had to be moved into a web system from existing data files. By hand, that would mean opening every record, transferring the required fields, checking the information and saving it in the system.

At just 2 minutes per record, the whole job would come to about 2,000 working hours, or more than 250 full working days.

At that scale, time is not the only problem. You need to know exactly which records have moved, which failed, which are missing data and which have to be retried.

What we built

We built an automated migration process that works through the records in sequence, validates the key fields and passes the data into the web system.

Every record carries a status: prepared, migrated, failed or needs review. Invalid and incomplete records are separated from the main flow, so a single bad record does not halt the whole process.

If the target system temporarily refuses data, the migration can be repeated for the failed records only.

What changed

Instead of a potential 2,000 hours of manual work, the 60,000 records are migrated automatically and human time goes only to exceptions and unclear cases.

At any point you can see how much data has moved, how much remains and why a particular record was rejected.

At this scale automation saves time and, just as importantly, keeps control and traceability from the first record to the last.

How the process worked before

Project information was spread across email, Excel files, chat apps and individual people’s notes. To understand the overall picture, a manager had to ask the team, check deadlines and gather the latest information by hand.

At an average of 5 minutes per project, a single pass across 25 projects takes more than 2 hours. Repeated several times a week, a significant share of a manager’s time goes into finding information rather than making decisions.

What we built

We built an internal project management system that shows active projects, owners, deadlines, current status and the main risks in one place.

The system can flag late work, approaching deadlines, projects without a recent update and situations that need a decision from management.

Each person updates only their own part of the process, and the overall picture assembles itself.

What changed

Instead of hours of information gathering, a manager can assess the whole portfolio in minutes and see immediately where attention is needed.

What matters most is that the system does more than display status. It creates one shared view of the work, so the team and the manager decide from the same information.

How the process worked before

The team had to agree about 30 contracts, acts and other documents each week. Several people reviewed each one, and some came back for corrections after comments.

Once files travel by email or chat, it quickly becomes unclear which version is current, who has already approved and whose response is still outstanding. Tracking documents alone came to about 5 hours a week.

Why SharePoint was not enough

The problem was not file storage. What needed managing was the whole approval flow: the owners, the different approval sequences, the deadlines, the returns for correction, the reminders and which version is in force.

What was needed was not another document repository, but a system to run the process around the document.

What we built

We built a system where each document is given an approval flow and its owners. The current status, the version in force, the approvals given and the next action are all visible.

Late approvals trigger reminders, and documents can be sent back for correction or routed down a different approval path.

What changed

Instead of about 5 hours spent tracking documents, the team spends roughly 1 hour reviewing exceptions and late cases.

A manager sees where every document stands in one place, and staff no longer hunt for the latest version or check who has yet to approve.

How the process worked before

Enquiries arrived by email, through website forms, social media and other channels. Staff forwarded them to one another, so it was not always clear who should reply or whether the customer had already been contacted.

At about 120 enquiries a week, even a few percent missed or forwarded late means lost sales opportunities. It was also hard for a manager to see quickly how many enquiries were waiting and where the process was stalling.

What we built

We built an internal enquiry management system that collects contacts from the different channels into one place and assigns them to the right person by defined rules.

Assignment can depend on the service, the region, the customer type, current workload or other criteria. Every enquiry has a status, an owner, a time of arrival and a next action.

If the customer is not contacted within the agreed time, the system reminds the employee or escalates to a manager.

What changed

The average handover time can drop from hours to minutes, and every contact has a clear owner from the moment it arrives.

A manager sees in one place how many enquiries came in, how many were handled, which are waiting and where the agreed response time has been exceeded.

The main change is that enquiries stop being scattered emails and messages and become a controlled sales process.

How the process worked before

Requests for leave, purchases, technical problems and other internal matters were sent by email, by message or handed straight to whoever was responsible.

At about 80 such requests a month the overall picture gets hard to see. Some items end up without a clear owner, people ask repeatedly about status, and a manager has to look through several channels for information.

What we built

We built an internal system where an employee picks the request type and provides the required information. Defined rules assign the request to the right person or send it for approval.

Every request carries a status, an owner, a submission date and a history of actions. Different request types can follow different flows: a purchase may need management approval, while a technical issue goes straight to the IT team.

What changed

Instead of several different communication channels, all internal requests are managed in one place.

An employee can see the status of their request, the owner knows what has to be done, and a manager can see how many requests came in, how many await a decision and where the work is piling up.

At 80 requests a month, saving even 5 minutes of admin on each adds up to more than 6 hours a month, before counting the time that used to go into repeated questions and searching for information.

How the process worked before

Service and technical work was scheduled through calls, messages and separate spreadsheets. A manager had to check by hand who was working that day, which jobs were already assigned, what was running late and where capacity remained.

With about 60 jobs scheduled a week, even a few changes a day cause confusion. One job overruns, another is cancelled, an urgent order arrives, and the plan has to be adjusted and communicated to the team again.

What we built

We built an internal scheduling system that shows all planned jobs, the staff responsible, the deadlines and the current status.

A manager can assign work by workload, skill, location or other criteria. Each employee sees their own assignments and updates the status, so the overall plan moves with the real situation.

Late, unfinished or unassigned jobs stand out immediately, and when an urgent order comes in it is quick to see where capacity is free.

What changed

Weekly scheduling and daily changes, which could take about 6 hours of a manager’s time, come down to roughly 1 hour a week.

The team sees the current plan in one place, and a manager can judge not only individual jobs but overall workload, delays and free capacity.

The system becomes more than a task list: it is the main tool for daily assignment and capacity planning.

How the process worked before

Waybills, freight documents and other logistics information arrived as PDFs, photos, scans and other formats. The documents were structured differently, so they could not be imported directly into the system in use.

Staff opened each document by hand, found the required fields and typed them into the system.

With, say, 5,000 documents a month at 3 minutes each, data entry alone comes to about 250 working hours a month, on top of the errors manual work brings and the extra checking they require.

What we built

We built an AI-based document processing flow that accepts logistics documents in different formats, recognises the document type and extracts the information the system needs.

The solution can extract the consignment number, dates, vehicle details, addresses, quantities and other defined fields. The data is validated against defined rules and prepared for transfer into the main system.

If a document is unreadable or the confidence is too low, it is passed automatically to a person for review.

What changed

The automated process reached 99.3% data recognition accuracy, so most documents are handled without manual entry.

Instead of hundreds of hours of data transfer, staff time goes only to the documents where the system finds something unclear or exceptional.

The main change is not only faster processing. A process that depended on constant manual entry became an automated, controlled part of the logistics system.

How the process worked before

Take a company that receives about 250 invoices a month from different suppliers. The documents arrive at different times, in different formats and not always to the same email address.

Someone has to find the invoice, download it, check the supplier and the period, file it correctly and pass it to accounting. At just 3 minutes per document, that comes to more than 12 hours of repetitive work a month.

What we built

The process monitors incoming email, recognises invoices and extracts the key document information. Each document is assigned to the right period, filed in the defined location and prepared for accounting.

Invoices can be sorted by supplier, month, company or other defined criteria. Where information is missing or a document cannot be recognised reliably, it is flagged for human review.

What changed

Instead of about 12 hours a month spent finding and sorting documents, roughly 1.5 hours goes into reviewing the prepared documents and the exceptions.

Most invoices travel the same defined path, and human time is left for the cases that genuinely need a decision.

How the process worked before

Suppliers sent quotes as PDFs, Excel files, email text or their own templates. One gave a total price, another a unit price, a third listed delivery, warranty or extra terms separately.

To compare 15 quotes, someone had to open each document, find the key information and copy it into a shared table by hand. At about 10 to 12 minutes per quote, preparing the data alone can take around 3 hours.

A further problem appears when the lowest price looks best but the quote differs on delivery time, warranty, quantity or another material term.

What we built

We built an AI-based quote analysis tool. It reads documents in different formats, extracts the key data and presents it in one consistent structure.

The AI can pull out price, timelines, quantities, warranty, payment terms and other criteria relevant to that purchase. It also flags mismatches, for example when a supplier offered a different equivalent product, a shorter warranty or left some services out.

Where the information is unclear or the quotes are not directly comparable, the case goes to a person for review.

What changed

Preparing 15 quotes for comparison drops from about 3 hours to 20 minutes.

Nobody retypes each quote by hand. They receive a normalised comparison and can spend the time on what matters: the choice itself and the risks.

The AI does not make the final decision. Its job is to read a large volume of information quickly, surface the differences and point to what deserves a closer look.

How the process worked before

Candidate information arrived from job boards, email, LinkedIn and other channels. CVs were structured differently, so the hiring team opened each document, found the required information and copied it by hand into a shared table or the recruitment system.

At an average of 8 minutes to review and enter one candidate’s information, 120 candidates make about 16 working hours of initial data preparation alone.

What we built

We built an AI-based process that collects candidate documents and prepares the information in one consistent structure.

From a CV the AI can extract professional experience, skills, education, languages, length of experience and other information relevant to that role. The candidate is assigned to the right opening, and missing or unclear data is flagged for human review.

The hiring team receives not a pile of CVs in different formats but uniformly prepared information that can be reviewed and compared quickly against the criteria for the role.

What changed

Preparing information on 120 candidates drops from about 16 hours to 2 hours of review and verification.

The AI does the repetitive collection and structuring work; the decision about a candidate still rests with a person.

Recruiters spend their time assessing experience and interviewing rather than copying CV data.