# AI Process Automation: Where to Start | Corex360 > AI automation starts with the job that repeats most: quote follow-up, contract renewal, leave approval. Which work an assistant runs alone, where it asks you. Source: https://www.corex360.com/en/blog/ai-automation-where-to-start Last updated: 2026-09-29 --- ![A row of pastel dominoes toppling in sequence, a copper-coloured piece at the front] Automation # Automating company processes with AI: where to start? AI automation is an arrangement that puts recurring work such as quote follow-up, contract renewal and leave approval under rules. Which jobs the assistant can run on its own and where it will ask you. [Request a demo](https://www.corex360.com/en/contact?konu=agents#form) [Corex360](https://www.corex360.com/en) [Blog](https://www.corex360.com/en/blog) Automating company processes with AI: where to start? Contents - 01 Which tasks suit automation? - 02 What does AI do beyond a rule? - 03 Which steps run automatically, and which wait for approval? - 04 Why does automation break when it is built in a separate tool? - 05 How does X AI run the process in Corex360? - 06 Where do you start with automation, and who is it too early for? Ş **[Şevket Erer](https://www.corex360.com/en/blog#blog-hakkinda)** Co-founder, AI Specialist · Corex360 22 September 2026 Updated 24 September 2026 5 min read [Modules](https://www.corex360.com/en/modules) [Solutions](https://www.corex360.com/en/solutions) Process automation with AI means tying the tasks that repeat the same way every week to a rule, and having an AI assistant read and carry out that rule. There is no need to start with a big project; start with a small task that breaks when someone forgets: an [unanswered quote](https://www.corex360.com/en/solutions/quote-tracking-software), a missed contract end date, a leave request lost in a message thread. ** 01. ## Which tasks suit automation? ** Short answer Tasks that repeat the same way, depend on a date or an event, and cause harm when forgotten suit automation; tasks that rely on judgement and negotiation stay with people. Example scenario: Picture a firm in İzmir that sells and installs air conditioning. A quote goes out, a reply is awaited, nobody follows up. After installation the service date comes round and the customer is not called. A leave request arrives on WhatsApp, the manager sees it and forgets. None of this is hard work; all of it depends on memory, and when the person changes, the process changes too. - Quote follow-up: the quote was sent, no reply came. - Contract renewal: the end date is near, no decision has been made. - Leave and advances: the request arrived, approval is pending. - Vehicles and service: a maintenance alert came in, no task was opened. - Incoming e-mail: a quote request, an invoice, or a complaint did not reach the right person. Look at the list: every line has a trigger (the quote went out), a condition (no reply came), and an action (send a reminder). That is the raw material of automation. Price negotiation, the phone call that saves a customer relationship, or a difficult conversation with an employee do not fit this pattern. There is a simple way to tell whether a task suits automation: look at how you would explain it to a new starter. If you can put it in one sentence, such as 'if no reply comes after the quote goes out, send a reminder', a rule can be written. If you would say 'use your judgement', the task should stay with a person. ** 02. ## What does AI do beyond a rule? ** Short answer A rule says what should be done; AI reads the incoming message, understands the subject, ties it to the right record, and prepares the draft. Classic automation works on neatly entered data: if the field is filled, the rule fires. In real life work is rarely entered neatly. A customer writes their quote request in the middle of a long e-mail, an invoice arrives as an attachment, a complaint hides inside a thank-you note. AI fills that gap. It reads the incoming text, finds the customer from the sender, and creates the record the rule can work on. You can read how a rule is read in [what a rule engine is](https://www.corex360.com/en/blog/what-is-a-rule-engine). The deciding point here is this: AI does not replace the rule; it prepares material for it. Back to the air-conditioning firm in İzmir. A building manager sends an e-mail: one of the units fitted last summer is dripping water, and meanwhile two more flats want air conditioning. One e-mail holds two jobs: a complaint and a quote request. AI files both on the same customer card and routes the complaint to its owner and the quote request to sales. ** 03. ## Which steps run automatically, and which wait for approval? ** Short answer At every step involving money, contracts, or people, the assistant prepares and asks; the manager decides. Which steps run automatically, and which wait for approval | Task | Step that runs automatically | Step that waits for approval | |---|---|---| | Unanswered quote | A reminder draft is prepared | The reminder goes with the salesperson's approval | | Shift left open | The system closes it at shift end and sends it to approval | The manager approves or rejects it | | Leave or advance request | The request reaches the manager through the approval flow | The manager approves it | | Incoming e-mail | The sender is matched to the customer card and the e-mail filed there | The owner writes the reply | | Discounts and payments | The information is gathered and submitted for approval | The authorised person decides | The main reason managers put off automation is the worry of losing control. Will the assistant send the quote on its own, grant the leave on its own? In a properly set-up system the answer is no. The assistant's job is to gather information, open the record, remind, and summarise; any promise made to a customer and any money leaving the company pass through a person's approval. The [guide on generative AI and personal data](https://www.kvkk.gov.tr/Icerik/8547/uretken-yapay-zeka-ve-kisisel-verilerin-korunmasi-rehberi-15-soruda) published by Turkey's Personal Data Protection Authority on 24 November 2025 also recommends that such systems be developed with a transparent, auditable, and human-centred approach. Every suggestion staying on the record is part of the trust as well. Who approved what, and when, is not searched for later; it is on the record. The structure that defines who passes an approval to whom is called an [approval chain](https://www.corex360.com/en/glossary#onay-zinciri). When a wrong suggestion can be undone, the team gets used to the assistant more easily. ** 04. ## Why does automation break when it is built in a separate tool? ** Short answer When the rule sits in one tool and the record in another, a change to either tool can stop the rule silently, and nobody notices. Many companies build automation outside the CRM, in a separate connector tool. The record is in the CRM, the rule in the connector, the result in an e-mail. For the first few weeks it works well. Then a field is renamed, an account password is renewed, and the rule stops working. The error message lands in an inbox nobody reads; the quote goes unanswered again, only this time everyone believes the automation is following it up. That is why the rule should sit next to the record. If the task it triggers, the customer it reminds, and the summary it produces all appear on the same card, a failure is noticed in the same place too. A quote with no task opened catches the eye of the first person who looks at its card. The same problem arises when the team changes. The person who built the rule leaves, and the connector account and its password go with them. If the rule is defined inside the record, it stays in place when people change, and the newcomer finds it on the same screen. ** 05. ## How does X AI run the process in Corex360? ** Short answer X AI reads e-mail and files it on the card, follows up rule-based tasks, and gives a morning report and an end-of-day summary; at sending, payment, leave, and permission steps it stops and asks. In Corex360, once company e-mail is connected, [X AI](https://www.corex360.com/en/x-ai) reads incoming e-mail, matches the sender to the customer card, and files it there. When a rule is broken, it opens a task and assigns an owner; it follows up a quote that has had no reply. It does not send a reply or a quote to a customer on its own; it prepares a draft, and sending happens with approval. X AI does not bring the team a new screen; it shows up next to the card as tasks, drafts, and reminders. You can see how it works step by step on [our page on handing the process to the assistant](https://www.corex360.com/en/solutions/hand-it-to-the-assistant). On approvals the line is clear: sending, discounts, payments, leave, and permission steps wait for your approval. Every morning X AI gives you one list of what it did, what is waiting, and what it is asking you; another summary follows when the day closes. ** 06. ## Where do you start with automation, and who is it too early for? ** Short answer Start with a single flow, which in most companies is quote follow-up; in a company where nobody can yet describe the process, automation is premature. The second thing that holds automation back is the idea of fixing everything first. The team tries to write down every process and then choose software; months pass and the repeating work is still done by hand. The better path is to start with a single flow. In most companies that flow is quote follow-up; once the habit settles, the second rule follows. A good candidate for the first rule is a task whose result shows straight away. If the list of unanswered quotes is shorter the next morning, the team trusts the rule. Starting with a rule whose result shows only months later lets interest fade before the habit takes hold. One more caution: automation does not fix a task that nobody can describe. If it is not clear who approves a quote and under which conditions, make that decision first. If you are looking for an assistant that decides entirely on its own and replies to customers by itself, X AI is not that; it prepares and asks. In the demo we pick your most repetitive task and write the first rule together. ## Sources - [Personal Data Protection Authority, guide on generative AI and personal data (in 15 questions)](https://www.kvkk.gov.tr/Icerik/8547/uretken-yapay-zeka-ve-kisisel-verilerin-korunmasi-rehberi-15-soruda) Last checked 24 September 2026 ## How was this page prepared? - Written by: [Şevket Erer](https://www.corex360.com/en/blog#blog-hakkinda), Co-founder, AI Specialist - Sources: legislation, official body, and vendor information is taken from the primary sources in the Sources list on this page; each source shows the date it was last checked. - Product information: sentences about Corex360 rest on the company's product fact records; unconfirmed features are not written. - Use of AI: the Corex360 team writes the text; AI tools are used for drafting and translation. [Publishing principles](https://www.corex360.com/en/about#yayin-ilkeleri) - Dates: first published 22 September 2026 , last updated 24 September 2026 . [Corrections policy](https://www.corex360.com/en/about#duzeltmeler) ## Frequently asked ** 01 What does business process automation with AI mean? ** It means defining a repeating task as a trigger, a condition, and an action, and having software carry out that definition. AI reads irregular input, such as an incoming message or document, and turns it into a record the rule can work on. ** 02 What is the difference between RPA and AI automation? ** RPA (robotic process automation) is software that imitates the clicks and copy-and-paste steps an employee makes on screen; it can break when the screen changes. AI automation reads the incoming text, understands the subject, and creates the record a rule can work on. RPA is mostly used to move data between older systems, while AI is used to turn irregular input into records. ** 03 Which department should automation start in? ** In the department where the task that repeats most, and loses a customer or money when forgotten, lives. In most companies that is sales: a quote goes out and a reply is awaited. Due-date follow-up in accounting and leave approval in HR are good starting points too; the department does not matter; the test is whether the task can be described in one sentence. ** 04 Are automation and workflow software the same thing? ** They are close but not the same. Workflow software maps the order of steps and who passes an approval to whom; automation ties the repeating steps to a rule and runs them without waiting for anyone. Where the two part is covered in our article on what a rule engine is. ** 05 Is it safe for the assistant to read our e-mail? ** The e-mail that is read is filed on the company's record, and access is role-based. In Corex360 data sits in a company-specific space on Corex360 Cloud or on your own server. Personal data is processed under KVKK (the Turkish Personal Data Protection Law). ** 06 How can you tell whether automation is working? ** Look at the list before any figures: are the lines for unanswered quotes, missed renewals, and pending approvals on the morning screen getting fewer, and does every line have an owner? If the rule is working, a problem shows up on the day it arises, without waiting for month end. ✳ **See this problem solved in your own work** We listen to your business with a few short questions and prepare a live online demo with your channels. [Request a demo](https://www.corex360.com/en/contact?konu=agents#form)[WhatsApp](https://wa.me/905321323288) Related articles ## Keep reading. [Back to the blog](https://www.corex360.com/en/blog) [Assistant What an AI assistant in a CRM does, and what it does not An AI assistant in a CRM reads, summarises, drafts, and follows up; the decision stays with you. Ask about the risk of wrong answers before you choose one. Date 23 September 2026 Read the article 01](https://www.corex360.com/en/blog/ai-assistant-in-crm)[Digitalisation Digitalisation in SMEs: where to start, why it stalls Digitalisation in SMEs means moving work records out of Excel, WhatsApp, and folders into one shared order. Why attempts stall, and which job to start with. Date 23 September 2026 Read the article 02](https://www.corex360.com/en/blog/sme-digitalisation)[On-premise What is on-premise CRM? What keeping data in-house really means On-premise CRM runs on the company's own server. Setup and updates then need a named owner, and security does not come by itself with the hardware. Date 23 September 2026 Read the article 03](https://www.corex360.com/en/blog/what-is-on-premise-crm)[Business intelligence What is business intelligence, and how does it differ from reports and KPI dashboards? Business intelligence links scattered records and brings forward what needs a decision. Its parts, how it relates to reports and KPI dashboards, and its uses. Date 22 September 2026 Read the article 04](https://www.corex360.com/en/blog/what-is-business-intelligence)[Business intelligence What is a KPI dashboard? The first screen a manager opens in the morning A KPI dashboard shows where a company stands against its goals, with thresholds and owners on one screen. Many get ignored over time, for reasons you can avoid. Date 23 September 2026 Read the article 05](https://www.corex360.com/en/blog/what-is-a-kpi-dashboard)[Business intelligence Management report or decision screen: where the Excel report stops A management report sums up a period for whoever decides; an Excel report freezes the moment it is made. Excel still does some of those jobs well. Date 23 September 2026 Read the article 06](https://www.corex360.com/en/blog/management-report-beyond-excel)[Assistant What is a rule engine in business processes: reminders, assignment, approval A rule engine turns work held in memory into a written rule: trigger, condition, action. In a sales office, some work fits a rule and approval stays. Date 23 September 2026 Read the article 07](https://www.corex360.com/en/blog/what-is-a-rule-engine)[Security Cloud or own server under KVKK: where does the data live Under KVKK (Turkish data law), cloud versus own server is less about location than about who processes and accesses data. That shapes what you ask. Date 23 September 2026 Read the article 08](https://www.corex360.com/en/blog/kvkk-cloud-vs-own-server) Was this page useful?