Automation services for business: measure first, then automate

Who we are

Tagula is an Israeli automation and AI services company operating from tagula.ai, serving companies and businesses across a range of sectors. We embed AI tools at every layer of the organisation: from organisational processes such as personal assistants on WhatsApp, through automation of repetitive office work, sales growth and lead generation, to complex control and oversight processes.

The work is done in Hebrew or in English, against the systems common in Israel, and by a fixed method: first measure how long the work actually takes, then simplify, and only then automate. The service is available across the country.

Behind the work is accumulated experience in technology, with an emphasis on artificial intelligence and model training, alongside experience in finance. The team includes AI engineers, economists, internal audit and information systems specialists, mathematicians and data scientists, among others. That is also what explains why the method opens with a measurement rather than a tool recommendation: people used to working with numbers do not propose a solution before seeing what the process actually costs.

What is business automation, and why it is not more software

Business automation takes repetitive work, the kind done the same way every week or every month, out of the hands of a person on your team. It moves that work into a process that runs without them. It does not replace your systems: it sits on top of them, pulls from them, cross-checks between them and returns a finished result to them.

Hence the confusion in search. The word "automation" on its own leads to industrial automation of production lines or to RPA tooling. Automation services for business are a different field: the repetitive office work that is done by exactly the people whose time costs the most.

The difference from buying another system is a question of measurement. Software is measured by what it can do. Process automation is measured by how much manual work disappeared after it went in. When the capabilities already exist and the problem is coordinating them, another system only adds a screen to fill in and another source of truth to contradict the last one.

This is what a candidate process looks like. A finance manager downloads a report from Hashavshevet, pastes it into a spreadsheet, cross-checks it against a table someone maintains by hand and emails it to management. That is not a finance manager's job, but the finance manager is the one doing it.

AI automation for business: what AI actually does here

AI does the part that rigid rules never could: reading free text, understanding what it says and turning it into a structured data point. Everything else, the calculation, the transfer, the production and the sending, stays deterministic, because in money and in legal documents nobody wants creativity.

The boundary is fairly clear. When the data sits in a defined field and only needs moving between systems, a direct connection between the interfaces does the job more cheaply and more reliably. AI comes in here: an email written in human language, a scanned document, a WhatsApp message, an invoice that every supplier sends in its own format.

The extreme case is CVs: dozens of documents per role, each in its own format, that someone reads by eye today. The full detail is on the Recruiting automation page. It covers daily screening from the mailbox and an old pool that proposes candidates for new roles. It also covers a list of who is ready to move, with no access to any platform.

Which repetitive work can be moved to process automation?

Four kinds of work repeat in almost every organisation, whatever the field: producing reports, generating documents, controls on exceptions and moving information between systems. From a business of three to an organisation of three hundred, the repetitive work looks the same. A monthly management report, hundreds of documents to their recipients and a deck built from raw data: these are examples from processes that have actually been built.

The report that writes itself

At the start of every month someone downloads files from several systems, lines up columns and pastes charts into a deck. Instead, the data is gathered from every source, cross-checked and delivered as a report or a formatted deck at a fixed time. Two days of teamwork become one minute, and the format stays the one management is used to. This is what it can look like in a typical organisation. Sales data from the core system, hours from the attendance system, project status in monday.com and expenses from the accounting system, in one file that arrives ready.

In the finance department this is the process that repeats the most, alongside bank and supplier reconciliations and document production. The full detail, including what Hashavshevet and Priority require and where AI is actually unnecessary, is on the Finance automation page.

Documents that go out on their own

The details change, the structure does not. A quote built from the data in the customer's record and sent out in a uniform design. A contract that fills itself in and goes out for signature. An order confirmation sent from WhatsApp Business, a renewal letter sent to a whole list. One or a thousand, the same effort.

An eye that does not miss

A price on an order that does not match the price list, a customer who has exceeded their credit limit, a supplier who sent the same invoice twice. Today these checks are done by eye, if at all, and always late. A continuous comparison between sources flags an exception the moment it happens, and not at month end, when we have already paid and it is already too late.

This is a whole field in itself, and our team has years of background in controls and internal audit. The full detail, including controls on payments, permissions and investments, is on its own page: Continuous controls in the organisation: catching what falls between departments.

Systems that talk to each other

No more double entry. For example: a customer opened in the CRM is opened in the accounting system too, with the same name and the same number. A form from the website opens a task and a folder in Microsoft 365. When a long-established Israeli system lives next to a new global SaaS tool, nobody designed them to work together, and that seam is exactly the work.

WhatsApp AI assistant for business: answering your customers 24/7

WhatsApp automation is an AI assistant that sits on your WhatsApp Business number and works in two directions. Outward, it holds a real conversation in Hebrew with whoever gets in touch and ends it with an action. An appointment booked, details collected, an enquiry passed to the right person. Inward, it is your own personal assistant: it takes tasks from you in the chat and carries them out. This is the channel Israeli customers actually use to contact a business. It is also the channel where the most enquiries get lost. A message that came in at 21:00 and was answered the next day at noon is a customer who has already called someone else.

Automated customer replies 24/7 do not mean replacing the human conversation. They mean that no message is left unanswered. When the conversation goes beyond what was defined, the assistant hands over to a person with all the context it collected, and does not guess.

Appointment-booking bot on WhatsApp

For hair salons, stylists, clinics and any business that runs a diary: the customer writes on WhatsApp, receives the open slots, chooses one and gets a confirmation. The appointment goes into the diary, a reminder goes out beforehand, and a cancellation frees the slot automatically. Anyone who manages appointments themselves knows it is never one appointment. It is dozens of exchanges a day that end in "when suits you".

Initial intake for law firms

A new enquiry arrives at an hour when nobody is available. The assistant collects what is needed to decide whether and how to handle it: the type of matter, the parties, the documents in hand, the timelines. Then it passes on an organised file. The lawyer joins the conversation when there is already something to read, instead of starting from "tell me what happened".

Intake is only the entrance. The repetitive work that really wears a law firm down starts after the file is opened.

A personal WhatsApp assistant for the business owner

This is the part most people do not think about: the assistant does not only answer your customers, it works for you. You write to it on WhatsApp the way you write to an employee. "Send Danny the quote", "remind me tomorrow to get back to Moshe", "what is the status of the Ramat Gan project". It does it. The interface is the chat that is open on your phone all day anyway: no extra app, no extra password, and no sitting at a computer.

What a personal WhatsApp assistant can do:

• Book and coordinate meetings: check the diary, propose times, send an invitation and update everyone when something moves.
• Send reminders, to you and to others: to get back to a customer, renew a licence, collect a payment, file a document before a deadline.
• Pull information from systems: project status, a customer's balance, inventory, a figure from the accounting system. A question in the chat instead of logging in to a system.
• Produce and send documents: a quote, an invoice, a confirmation or a contract sent to the right recipient without you opening a file.
• Record and log: a voice message you recorded on the road becomes a call summary, a task or a note on the customer's record.
• Summarise your day: what came in, what is urgent, what is waiting for your answer, in one message in the morning or the evening.
• Alert on exceptions: a customer who has not paid, an enquiry that was not answered, a deadline approaching.
• Pass tasks on: open a task for an employee, assign an owner and make sure it was closed.

This is especially useful for anyone who does not sit at a computer. A manager who is on site, a tradesperson between one customer and the next, or a manager whose day is mostly meetings.

Law firm automation: the work nobody wants to do

In a law firm the repetitive work does not sit in the interesting cases but around them. It is done in hours that could have gone to the law itself. These are the processes that repeat in every firm, whatever the practice area:

• Filling in recurring documents. Powers of attorney, routine statements of claim, motions, agreements with a fixed structure and affidavits. The details change, the skeleton does not. The document fills itself in from the file's data and comes out ready for review. It is not opened from a previous version for another client, with all the risk that carries.
• Complex calculations. Interest and indexation, severance pay calculations, dividing an estate, debts under debt enforcement, fines and fees. These are calculations done in a private Excel someone built once, where every mistake is expensive: exactly the kind of work a computer gets right every time.
• Deadline tracking. Filing dates, expiring limitation periods, hearing dates and client reminders. A missed deadline is the most expensive professional risk in a firm, and today it is kept in a diary and in memory.
• Scanning and extracting information from documents. A scanned contract, a bank statement, a court ruling: pulling the relevant fields into the file, instead of typing from paper.
• Time reports and billing. Consolidating the hours and the file's expenses and producing an invoice, including reconciliation with the accounting system.
• Client updates. The "what is happening with my case" message that comes back again and again, and can be answered automatically from the real status.

The boundary stays clear: the system prepares, calculates and organises, and legal judgement and the signature stay with the lawyer. We do not build a substitute for legal advice. We take off the desk what eats hours and does not require a licence. The full detail, including deadline tracking and what happens to client material, is on the automation for law firms page.

Lead generation and marketing automation: so growth stops depending on spare time

In most businesses marketing is the first thing to drop when the week is full. Not because it does not matter, but because it is the only thing nobody shouts about. Automation puts it back into action even in the weeks when nobody has time for it.

• Lead sourcing by your line of business. Instead of a generic list, ongoing mapping of businesses or projects that match exactly what you sell. The criteria are the ones you define: industry, size, region, project stage or an event that signals a need.
• Enrichment and prioritisation. A lead that comes in does not stay a name and a phone number. The system fills in information about it from open sources and ranks it, so your time goes first to whoever is likely to close.
• An immediate reply to a new enquiry. A lead that gets an answer in minutes rather than the next day is a lead that is still talking to you and not to whoever answered first. This is the direct link to the WhatsApp assistant above.
• Follow-up that does not slip. Who has not replied, who is due a reminder, which quote has been sitting for two weeks without a decision. That continuity is usually the difference between a deal that closes and a deal that is forgotten.
• Marketing content that goes out on schedule. A newsletter, an update to existing customers, a post or a professional document. Built from the material you already have and sent on a timetable, instead of when a moment frees up.
• Measurement that shows what worked. Which source the enquiries that were really worth it came from, so the decision on where to invest rests on data and not on a feeling.

The goal is commercial and clear: more relevant enquiries, fewer enquiries that fall through the cracks, and marketing that keeps running in a busy month. What it does to your revenue depends on the market, on price and on what happens after the lead comes in. So it is measured at yours and not promised in advance.

Which systems it connects to, and what happens when a system does not cooperate

The honest answer is that it depends on the system, and that it is checked before anything is promised. Some systems have a proper interface we work against directly. Some allow only file export and import, or scheduled reports. And some actions the vendor has not opened to outside access.

The typical Israeli environment is made up of a finance system like Hashavshevet, a core system like Priority and invoicing in Green Invoice. Add work management in monday.com, email and files in Microsoft 365, and customers who close deals on WhatsApp Business. Checking the connection against the system itself is part of the measurement stage: in the version and with the permissions you have, not according to what it is said to allow. Promising an integration in a sales call, without opening the system, is the fastest route to a project that stalls halfway.

The most common pair is the CRM and the invoicing system. What crosses between them, what stays manual, and the three failure points that break such a connection are set out on the CRM to invoicing sync page.

What does working with us look like, step by step?

Three stages, in a fixed order: we measure how long the work actually takes, then we build one process and run it alongside the existing work. Then we hand it over to you working. A stage does not start before the one before it has proven itself.

01 Measure: business automation consulting starts with numbers, not with a tool recommendation

Consulting that opens with a tool recommendation is a guess. So we sit with whoever runs the process today and measure how long it actually takes, not how long it feels. That means sitting next to the collections manager while the letters go out, counting actions, and asking what happens when a customer does not appear in the system. By the end of this stage you know which process is worth the most and how much it is expected to return. So far we have not touched a single system. The whole method, including the four numbers that get measured and the five questions that set the order, is on the process mapping for automation page.

02 Build and run: you watch it work before you rely on it

We start with one process, the one with the highest return. It runs alongside the existing work until the results match, so you test it against reality rather than against a promise. If the results do not match, we fix it. There is no moving to the next stage because a date arrived.

03 Hand it over: it works where you are, day to day

The solution enters your routine with full training and a process owner on your side. That is one person who knows the work, knows what a correct result looks like, and knows how to start and stop it. From there it simply runs, and we stay available at whatever level you choose. That order is also an order of priorities: we measure, we simplify, and only then do we automate.

When not to automate a business process, and when we say no

Business automation does not suit every process, and some processes are better cancelled than accelerated. Automating an unnecessary process does not remove it. It only makes it happen faster, at a fixed maintenance cost, and without anyone asking again whether it is needed at all.

Some measurements end with the conclusion that the saving does not justify the investment, and that is a legitimate answer. These are the cases where we will tell you no, even when you ask for yes:

• A report nobody reads. Before we speed up the production of a report, we check who opens it and what they do with the numbers. A report nobody opens does not need to be produced faster. It needs to stop being produced.
• A process that exists to compensate for a fault somewhere else. If a figure is typed twice because one field is not synchronised, the fix is in the field. Automating the double entry preserves the fault forever.
• A process whose rules have not been agreed yet. When every department keeps its own version of the same table, there is no technical problem here but a lack of agreement on who owns the number. Connecting the systems only makes the disagreement faster.
• Input nobody fills in. If the field in the system stays empty whenever filling it is inconvenient, there is nothing to cross-check and nothing to alert on. First the input is sorted out.
• A decision that is entirely judgement. Approving an exceptional discount, extending credit to a new customer, deciding to part with a supplier. All the material for the decision can be prepared, but a person signs the decision.
• A process that is about to disappear. If you are in the middle of replacing a core system, building connections to the outgoing one is a waste. We wait.

This list is the reason the first stage is measurement and not a quote. A vendor who agrees to every request has already decided what they are selling you, before seeing what you have.

Who is business process automation for?

Automation services for business suit any organisation where the same work repeats at a steady pace, from a business of three to an organisation of three hundred. What changes between the two ends is who absorbs the manual work, how many people have to say yes, and what makes the project hard.

In a small business

The owner is the bottleneck. The same person sends quotes, chases payments, issues invoices in Green Invoice and coordinates on WhatsApp, and in between is supposed to sell. When the process is not written down anywhere, the measurement stage is also the first time it is documented. The decision is quick, because the person deciding is also the person suffering. And the time it frees is the time that makes money: the hours that come back are selling hours.

In a large organisation

The repetitive work is split between many people, so it is almost invisible. The pain sits in the seams. A handover between departments, reconciling the core system with a spreadsheet someone maintains on the side, a report assembled from sources that do not agree with each other. The decision is slower, and that is exactly why the measurement stage is critical: it is what turns a feeling into a budget request that can be defended. The parallel run matters especially here too, because nobody moves a financial process to a new track on the strength of a demo.

At both ends we start with one process, and only after it do we talk about the next.

By industry

In some industries the repetitive work is so typical that it can be described in advance:

• Law firms: initial intake from new enquiries, producing recurring documents, tracking deadlines.
• Hair salons, stylists and clinics: booking appointments on WhatsApp, reminders, and filling slots that opened up.
• Insurance agents: renewals, comparing data between sources and alerting on gaps.
• Contractors and field service providers: quotes, customer approvals, and status tracking with end customers.
• Architects and consultants: professional opinions and reports that repeat in a fixed structure.

What happens to your data

A customer's WhatsApp enquiry, a scanned document and a financial report are sensitive business data. So this is a question we talk about before we start, including the part it is convenient not to say:

• A process that uses AI sends data to a third-party provider. The provider may collect, store or use it. Each provider has its own policy, set by the provider and changed from time to time, and we neither control it nor answer for it. Anyone who promises you otherwise is promising something that is not in their hands.
• So first we map what has to go out at all. Before a tool is chosen, we check what data the process must see and what can stay out. The less data leaves, the less depends on someone else's policy.
• There is also a path that never leaves the organisation. When the sensitivity justifies it, a tool that runs on your own infrastructure can be considered. That is an option checked case by case (including cost and performance), and not the default.
• Minimum permissions. The access we ask for a process is what it needs to work, no wider than that.
• What is binding is in the agreement. A marketing page is not where what is kept, where and for how long gets decided. That is set in a written agreement with you, after it is known exactly what is involved.

The decision on what data enters the process at all is yours. A business that handles medical or legal data does not get the same answer as a business that books appointments, and that is discussed at the measurement stage.

What it looks like at clients

Examples from actual work, at different ends of the same method:

• Metal Stonespecialist work at height
• Negishimaccessibility consulting for buildings and infrastructure
• QTimepersonal companionship for seniors

Metal Stone: specialist work at height

The company's managers are on site most of the day, on roofs and on scaffolding, with no access to a computer. The information they need to make a decision sits elsewhere: data from the accounting system and information from the company's systems. And the customer standing in front of them needs a quote. Now, not in the evening from the office.

What was built:

• Access to data from the field: information from the accounting system and the company's systems is available from the phone, without going back to the office.
• Quotes produced in a short time, in front of the customer, instead of promising "I will send it tomorrow".
• Real-time alerts on employee training and refreshers: in an industry where an expired certification is both a safety risk and a regulatory exposure, not a diary item.
• Smart inventory management: what there is, where, and when to reorder.
• A customer area: every customer logs in with their own account and sees only their own data. Project status, quotes to approve or decline on the spot, invoices and documents, and submitting enquiries with files.

What changed: The company reports a saving of dozens of hours a month in producing quotes. Beyond the time, what can be done with it changed too. A project can be looked at in depth and priced more professionally, instead of a quick price going out because there is no time to check.

I'm on site most of the day, not at a computer. A customer used to ask for a price and I'd tell him "I'll send it tomorrow evening". Today I send a quote on the spot, from my phone. And apart from the time it saves, I can finally look at the project in depth and price it right, instead of throwing out a number because I have no time to check.

Ohad, Metal Stone

Negishim: accessibility consulting for buildings and infrastructure

The work of a licensed accessibility consultant is hard professional work: reviewing architectural plans, checking compliance with regulations and standards, and forming a professional opinion its author can stand behind. Around that professional core, whole processes had built up that were done by hand, and those are what we moved to automation. The time that was freed went back to the professional work itself.

In the work of an architect and a consultant, these are the processes that repeat and can be automated:

• Producing professional opinions and surveys: a fixed document structure with changing content, built from the findings collected on site instead of by copying a previous version.
• A regulations checklist: cross-checking the findings against the requirements of the relevant standard, so a clause is not dropped because of tiredness at the end of the day.
• Extracting information from plans and bills of quantities: pulling the relevant dimensions and data into the report.
• Documenting site visits: photos and notes from the phone that go into the report in the right place, instead of sorting them in the evening.
• Tracking with authorities and clients: where every file stands, what was sent and what is waiting to be completed.
• Quotes and billing by project scope and type of inspection.

What changed: The firm reports that the analysis and review step of a plan used to take between 5 and 20 hours per plan, and now runs in about 5 minutes. The time that was freed went back to the professional work itself.

Here too the boundary holds: the system prepares and cross-checks, the professional opinion and the professional responsibility stay with the licensed accessibility consultant.

Analysing a plan used to take me between five and twenty hours, depending on its size. Today that part runs in a few minutes and I get the findings laid out in front of me. What mattered to me was that the opinion stays mine: the system prepares and cross-checks, the professional decision and the responsibility are mine. The time that was freed went exactly there.

Yuval, Negishim

QTime: personal companionship for seniors

QTime provides personal companions who accompany seniors in activities at home and outside it: sport, brain training, learning technology, outings and accompaniment on a day of medical tests. Every activity is a coordination between a person, a companion, a place and a time, and that coordination fell entirely on the manager's diary.

What was built: a personal WhatsApp assistant that runs the day from the chat, without opening a system and without sitting at a computer.

What changed: Management reports a saving of at least around two hours a day in managing personal time.

My business involves dealing with customers over and over. It all fell on my diary. Today I write a WhatsApp message and it happens. It gives me back at least two hours a day, and I'd rather spend them on people than on the diary.

Yuval Gabay, QTime

What is needed from you for this to work?

Three things, and without them it does not work. First, time from the person who actually does the work, for the measurement. Second, access to the systems the process has to touch. Third, a process owner on your side who keeps holding it after the handover. A programmer is not needed to run the process day to day.

Where to start?

Pick one process that repeats itself and annoys you, and call or send an email. From there the measurement stage begins: we sit with whoever runs the process today and check how long it actually takes. At the end you have a priority order and an estimate of how much each process is expected to return.

It helps to arrive with three details: who does the process, how often, and where the information it consumes comes from. On our side is Avishai, Tagula's founder.

• Email: info@tagula.ai
• Phone and WhatsApp: 054-650-4053

Less work, more growth.

Frequently asked questions