{"id":77155,"date":"2026-01-02T16:19:00","date_gmt":"2026-01-02T10:49:00","guid":{"rendered":"https:\/\/www.tothenew.com\/blog\/?p=77155"},"modified":"2026-01-02T18:13:35","modified_gmt":"2026-01-02T12:43:35","slug":"exploring-agentic-ai-for-automation","status":"publish","type":"post","link":"https:\/\/www.tothenew.com\/blog\/exploring-agentic-ai-for-automation\/","title":{"rendered":"Exploring Agentic AI for Automation"},"content":{"rendered":"

The other day, my client asked for something that sounded simple at first – he wanted an alert in his inbox whenever the date changed on a specific website.
\nAnd, of course, the first thought that came to my mind was: \u201cYes, that\u2019s doable with Python!\u201d
\nBecause that\u2019s what I have always done in my analytics journey – write code to automate stuff.<\/p>\n

So I rolled up my sleeves and started working on it. The plan was straightforward: check if the website had any security or login requirements, scrape the latest date, compare it with the previous one, and email myself if it changed.<\/p>\n

Since I couldn\u2019t expose my Gmail password, I had to create an OAuth token for sending the alert email securely – a small but necessary step. And yes, after a few tweaks, the code worked perfectly.<\/p>\n

import os<\/em>\r\nimport json<\/em>\r\nimport smtplib<\/em>\r\nimport re<\/em>\r\nimport cloudscraper<\/em>\r\nimport re<\/em>\r\nfrom bs4 import BeautifulSoup<\/em>\r\nfrom datetime import datetime<\/em>\r\nfrom email.mime.text import MIMEText<\/em>\r\nfrom selenium import webdriver<\/em>\r\nfrom selenium.webdriver.chrome.service import Service<\/em>\r\nfrom selenium.webdriver.common.by import By<\/em>\r\nfrom selenium.webdriver.support.ui import WebDriverWait<\/em>\r\nfrom selenium.webdriver.support import expected_conditions as EC<\/em>\r\nfrom webdriver_manager.chrome import ChromeDriverManager<\/em>\r\n\r\n# Configurations<\/em>\r\nURL = \"https:\/\/echa.europa.eu\/news-and-events\/news-alerts\/archive-2025\"<\/em>\r\nSTATE_FILE = \"echa_news_state.json\"<\/em>\r\nEMAIL_FROM = \"xxx@tothenew.com\"<\/em>\r\nEMAIL_TO = \"xxx@tothenew.com\"<\/em>\r\nEMAIL_PASS = \u201cccccc\"<\/em>\r\nSMTP_SERVER = \"smtp.gmail.com\"<\/em>\r\nSMTP_PORT = 587<\/em>\r\n\r\ndef get_latest_news_date():<\/em>\r\n  \u00a0 url = \"https:\/\/echa.europa.eu\/news-and-events\/news-alerts\/archive-2025\"<\/em>\r\n  \u00a0 scraper = cloudscraper.create_scraper() \u00a0<\/em>\r\n  \u00a0 response = scraper.get(url)<\/em>\r\n  \u00a0 response.raise_for_status()<\/em>\r\n  \u00a0 soup = BeautifulSoup(response.text, \"html.parser\")<\/em>\r\n  \u00a0 date_pattern = re.compile(r\"\\b\\d{1,2}\\s+\\w+\\s+2025\\b\")<\/em>\r\n  \u00a0 match = date_pattern.search(soup.get_text(\" \", strip=True))<\/em>\r\n  \u00a0 if match:<\/em>\r\n  \u00a0 \u00a0 \u00a0 return match.group()<\/em>\r\n  \u00a0 else:<\/em>\r\n  \u00a0 \u00a0 \u00a0 return None\r\n<\/em>\r\ndef load_previous_state():<\/em>\r\n  \u00a0 if os.path.exists(STATE_FILE):<\/em>\r\n  \u00a0 \u00a0 \u00a0 with open(STATE_FILE, \"r\") as f:<\/em>\r\n  \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 return json.load(f)<\/em>\r\n\u00a0 \u00a0 return {}<\/em>\r\n\r\ndef save_state(state):<\/em>\r\n  \u00a0 with open(STATE_FILE, \"w\") as f:<\/em>\r\n\u00a0 \u00a0 \u00a0 \u00a0 json.dump(state, f)<\/em>\r\n\r\ndef send_email(latest_date, previous_date):<\/em>\r\n  \u00a0 subject = \"\ud83d\udce2 ECHA News Alert Updated\"<\/em>\r\n  \u00a0 body = f\"\"\"<\/em>\r\nLatest news date from ECHA Archive 2025: {latest_date}<\/em>\r\nPrevious recorded: {previous_date or 'None'}<\/em>\r\nURL: {URL}<\/em>\r\n\"\"\"<\/em>\r\n  \u00a0 msg = MIMEText(body, \"plain\")<\/em>\r\n  \u00a0 msg[\"From\"] = EMAIL_FROM<\/em>\r\n  \u00a0 msg[\"To\"] = EMAIL_TO<\/em>\r\n  \u00a0 msg[\"Subject\"] = subject<\/em>\r\n\r\n  \u00a0 with smtplib.SMTP(SMTP_SERVER, SMTP_PORT) as server:<\/em>\r\n  \u00a0 \u00a0 \u00a0 server.starttls()<\/em>\r\n  \u00a0 \u00a0 \u00a0 server.login(EMAIL_FROM, EMAIL_PASS)<\/em>\r\n\u00a0 \u00a0 \u00a0 \u00a0 server.send_message(msg)<\/em>\r\n\r\ndef main():<\/em>\r\n  \u00a0 latest_date = get_latest_news_date()<\/em>\r\n  \u00a0 if not latest_date:<\/em>\r\n  \u00a0 \u00a0 \u00a0 print(\"\u26a0\ufe0f Could not find a date on the page.\")<\/em>\r\n\u00a0 \u00a0 \u00a0 \u00a0 return<\/em>\r\n\r\n  \u00a0 state = load_previous_state()<\/em>\r\n\u00a0 \u00a0 previous_date = state.get(\"latest_news_date\")<\/em>\r\n\r\n  \u00a0 if latest_date != previous_date:<\/em>\r\n  \u00a0 \u00a0 \u00a0 send_email(latest_date, previous_date)<\/em>\r\n  \u00a0 \u00a0 \u00a0 state[\"latest_news_date\"] = latest_date<\/em>\r\n  \u00a0 \u00a0 \u00a0 save_state(state)<\/em>\r\n  \u00a0 \u00a0 \u00a0 print(f\"\u2705 Email sent. New date: {latest_date}\")<\/em>\r\n  \u00a0 else:<\/em>\r\n\u00a0 \u00a0 \u00a0 \u00a0 print(f\"No change. Latest date still: {latest_date}\")<\/em>\r\n\r\nif __name__ == \"__main__\":<\/em>\r\n\u00a0 \u00a0 main()<\/em><\/pre>\n
Then came the challenge: automation.
\nI explored multiple approaches – Windows Task Scheduler, Cron Jobs, even a few GitHub scripts. I ended up using Task Scheduler because I didn\u2019t have access to a cloud platform. The setup worked, but there was a catch – my system had to be on at the scheduled time.
\nStill, I was happy. I thought, \u201cAt least it works!\u201d<\/p>\n
I demoed it to my client, and he smiled – and then rejected the idea.
\nNot because it didn\u2019t work, but because the intended end-user wasn\u2019t technical. The client wanted someone with zero coding knowledge to handle this task.<\/p>\n
Then came the real turning point.
\nHe asked me, \u201cHave you explored Agentic AI?\u201d<\/p>\n
My first thought was, \u201cCome on, we can\u2019t do this using ChatGPT – how would we even schedule it?\u201d
\nBut curiosity got the better of me, and that\u2019s when my two-week coding project turned into a two-day discovery of what\u2019s possible today.<\/p>\n
What is Agentic AI, really?<\/strong><\/h2>\n
Agentic AI refers to AI systems that can take autonomous actions to achieve goals, not just answer questions.
\nUnlike traditional AI chatbots that stop at providing text responses, agentic systems can plan, reason, execute, and monitor workflows – almost like a digital assistant that doesn\u2019t just tell you what to do but actually does it.<\/p>\n
Think of them as a bridge between LLMs (like GPT-4\/5) and automation tools – they combine reasoning with execution. They can log into websites, extract information, perform checks, send alerts, or even trigger workflows without you having to write a line of code.<\/p>\n
In my case, instead of coding a scraper, an AI agent could watch a web page for me and email me when something changes. That\u2019s exactly what I found in Browse.ai.<\/p>\n
My hands-on with Browse.ai<\/strong><\/h3>\n
At first, I was skeptical – \u201cthere\u2019s no way this thing will actually detect a date change and send an email automatically.\u201d
\nBut once I started exploring, the simplicity was shocking.<\/p>\n
All I did was:<\/p>\n
    \n
  1. Go to Browse.ai<\/a><\/li>\n
  2. Select the action you want to perform like Monitor Site Changes or Extract Structured Data from a website<\/li>\n<\/ol>\n
    \"image\"<\/a>
    nn<\/p><\/div>\n
    3. Select Monitor Site Changes and paste the website link (which requires monitoring),<\/p>\n
    \"image\"<\/a>
    image<\/p><\/div>\n
    4. Now once the web page uploads on screen, select either Text or capture screenshot to train the model<\/p>\n
    \"image\"<\/a>
    image<\/p><\/div>\n
    5.Now Highlight the text (in my case date and List name)<\/p>\n
    \"image\"<\/a>
    image<\/p><\/div>\n
    6.Now Name your Robot<\/p>\n
    \"image\"<\/a>
    image<\/p><\/div>\n
    7.Set a daily\/weekly or monthly schedule,<\/p>\n
    \"image\"<\/a>
    image<\/p><\/div>\n
    8.And choose email notification.
    \nAlso, we can view all Trained Robots at home page for easy navigation<\/p>\n
    \"image\"<\/a>
    image<\/p><\/div>\n
    That\u2019s it – no scripts, no tokens, no schedulers, nothing.<\/p>\n
    And then, a few days later, I got an email from Browse.ai: \u201cChange detected on the monitored page.\u201d<\/h4>\n
    How the actual Output Looks like:<\/h3>\n
    \"Final<\/a>
    Final Output<\/p><\/div>\n
    That was my \u201cwow\u201d moment. What took me days to code, schedule, and secure could be done in less than 15 minutes.<\/p>\n
    About Browse.ai (Free vs Paid)<\/h3>\n
    Browse.ai works on a credit (token) system.<\/p>\n
    When you sign up, you get 50 free tokens, enough for testing 50 tasks (like 50 page runs or extractions).
    \nEach run or monitoring cycle consumes a few tokens depending on what you\u2019re doing (a simple date extraction uses 1 token).
    \nYou can schedule robots to run daily, weekly, or hourly – the frequency and number of monitored pages determine how fast you use your tokens.<\/p>\n
    \u00a0Free Plan:<\/strong>
    \n1. 50 free credits (tokens)
    \n2. Basic data extraction
    \n3. 1 monitoring robot
    \n4. Manual and daily scheduling<\/p>\n
    Paid Plans (as of now):<\/strong>
    \n1. Start from around $39\/month
    \n2. Include 2,000+ credits
    \n3. Allow multiple robots, advanced scheduling, webhooks, and API access (so you can connect with Slack, Zapier, or Gmail).<\/p>\n
    For most business cases, the free version is enough to test the workflow or set up a single monitor. Once you scale or need multiple pages tracked, you can move to the paid plan.<\/p>\n
    What I learned<\/strong><\/h2>\n
    Exploring Browse.ai and understanding Agentic AI made me realize how automation is evolving. It\u2019s not just about writing scripts anymore,\u00a0 it\u2019s about letting AI act for you.
    \nAgentic systems like Browse.ai, Zapier\u2019s AI agents, or even OpenAI\u2019s upcoming \u201cProjects and Actions\u201d features represent a shift from prompting AI to delegating work to AI.<\/p>\n
    So yes, the two weeks I spent coding were not wasted, they helped me appreciate the simplicity of these tools even more.
    \nAnd now, if you ask me to set up a page-change alert for a non-technical client, I know exactly what to recommend.<\/p>\n
     <\/p>\n
     <\/p>\n","protected":false},"excerpt":{"rendered":"
    The other day, my client asked for something that sounded simple at first – he wanted an alert in his inbox whenever the date changed on a specific website. And, of course, the first thought that came to my mind was: \u201cYes, that\u2019s doable with Python!\u201d Because that\u2019s what I have always done in my […]<\/p>\n","protected":false},"author":1789,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"iawp_total_views":52},"categories":[5879],"tags":[4782],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/posts\/77155"}],"collection":[{"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/users\/1789"}],"replies":[{"embeddable":true,"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/comments?post=77155"}],"version-history":[{"count":6,"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/posts\/77155\/revisions"}],"predecessor-version":[{"id":77293,"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/posts\/77155\/revisions\/77293"}],"wp:attachment":[{"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/media?parent=77155"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/categories?post=77155"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.tothenew.com\/blog\/wp-json\/wp\/v2\/tags?post=77155"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}