On September 1, 2026, Loftal unveiled Pigeon Workflow, a next‑generation AI automation platform that promises to move far beyond the fragile scripts of traditional RPA. By combining visual AI perception with large‑language‑model reasoning, the tool lets anyone design end‑to‑end browser‑based processes without writing a line of code. This launch arrives at a moment when enterprises are under pressure to squeeze more productivity from existing staff while avoiding the costly downtime that brittle bots cause when a web page changes its layout. Loftal’s pedigree—backed by an AI‑focused award from ASPIC and a track record with its PigeonCloud database—gives the new product credibility in a crowded market. In the following sections we explore how the platform tackles the core weaknesses of RPA, what practical benefits it delivers, and why it may signal a shift toward truly autonomous digital workers.
Traditional RPA relies on hard‑coded selectors such as DOM IDs or XPaths to interact with web applications. While this approach works well for stable internal systems, it falters the moment a front‑end team tweaks a button’s class or redesigns a page layout, causing the bot to throw an exception and halt the entire workflow. The resulting maintenance overhead—often measured in hours per week per bot—erodes the promised savings and forces organizations to keep a dedicated RPA support team on standby. Moreover, RPA lacks any capacity for judgment; it cannot interpret a vague email request, decide which invoice matches a purchase order, or generate a thoughtful reply to a customer query. These limitations confine RPA to pure copy‑paste tasks and prevent it from handling the knowledge‑intensive steps that dominate modern office work. Pigeon Workflow sidesteps these issues by replacing selector‑based interaction with visual perception, allowing the AI to locate elements by appearance rather than brittle code.
The visual AI engine at the heart of Pigeon Workflow continuously scans the browser viewport, using convolutional neural networks trained on thousands of UI patterns to recognize buttons, fields, menus, and dialogs regardless of underlying HTML changes. When a target element moves or is restyled, the model still identifies it by its visual cues—color, shape, iconography, and relative position—so the automation continues uninterrupted. This approach dramatically reduces the failure rate associated with UI updates; early beta users reported a drop from weekly interventions to less than one incident per month across dozens of bots. Additionally, the visual layer enables cross‑platform consistency: the same workflow can run on Chrome, Edge, or Firefox without needing separate selector sets. By decoupling automation logic from the underlying markup, Loftal delivers a more resilient foundation that scales with the rapid UI iterations typical of SaaS applications and internal portals.
Beyond stable interaction, Pigeon Workflow equips each workflow step with access to a library of over fifty AI nodes powered by state‑of‑the‑art large language models. These nodes can perform tasks that previously required human cognition: extracting key qualifications from a resume and generating tailored interview questions, summarizing lengthy meeting transcripts into actionable bullet points, drafting polite yet precise email replies based on the tone and content of an incoming inquiry, or even classifying support tickets according to urgency and topic. Because the LLM nodes are prompted with contextual data from earlier steps—such as the extracted text from a form or the sentiment of a chat message—they produce outputs that are both relevant and coherent. This capability transforms the platform from a simple macro recorder into a true decision‑support system, enabling businesses to automate processes that involve judgment, language understanding, and creative synthesis.
Loftal built Pigeon Workflow on the same drag‑and‑drop canvas that made its PigeonCloud database popular among non‑technical users. The interface invites analysts, office managers, and citizen developers to record a sequence of browser actions—clicks, typing, file uploads—then enhance each step with an AI node from the sidebar. Connecting nodes is as simple as dragging a line between output and input ports, instantly defining the data flow. Because the platform abstracts away syntax, loops, and error handling, a complex end‑to‑end process that might take a junior developer days to script in traditional RPA can be assembled in under fifteen minutes by a business analyst with no coding background. The low barrier to entry encourages rapid experimentation, allowing teams to prototype improvements to invoicing, employee onboarding, or customer feedback loops without waiting for IT sprint cycles.
The intelligent automation market is projected to exceed $30 billion by 2028, driven by a convergence of RPA, AI, and low‑code platforms. Analysts note that the fastest‑growing segment is AI‑augmented automation, where bots gain perception, language, and decision‑making abilities. Loftal’s entry into this niche is timely: enterprises are seeking solutions that not only cut repetitive work but also adapt to evolving digital experiences without constant retooling. Pigeon Workflow’s pricing—zero upfront fee and a monthly subscription starting at ¥50,000 (≈$330)—places it within reach of mid‑size departments while still offering enterprise‑grade scalability. Compared with legacy RPA licenses that often run into six‑figure annual costs per bot, the subscription model reduces capital expenditure and aligns spending with actual usage, a factor that procurement teams increasingly favor in tight budget environments.
Loftal competes in a space occupied by established players such as UiPath, Automation Anywhere, and Blue Prism, all of which have begun adding AI computer vision and LLM modules to their offerings. However, many of these extensions remain separate add‑ons that require additional licensing, specialized training, and complex integration projects. Pigeon Workflow differentiates itself by delivering vision and language capabilities as core, tightly coupled features within a single no‑code environment. Newer entrants like Microsoft’s Power Automate with AI Builder and Google’s AppSheet also provide low‑code AI flows, but they often rely on proprietary cloud services that can raise data‑privacy concerns for organizations with strict sovereignty requirements. Loftal’s on‑premise‑friendly deployment options and transparent pricing give it an edge for industries such as finance, healthcare, and government where control over data handling is paramount.
Consider a typical HR onboarding workflow: a new hire’s offer letter arrives via email, the system extracts personal data, populates a web‑based onboarding portal, schedules orientation sessions, and sends a welcome package. With Pigeon Workflow, an HR coordinator can record the steps of logging into the portal, filling in fields, and uploading documents, then attach an LLM node that reads the offer letter and auto‑fills missing information such as emergency contacts or banking details. Another node can generate a personalized welcome email based on the hire’s role and location. In finance, the platform can monitor a supplier invoice portal, capture line‑item data, cross‑check it against purchase orders in an ERP system, and flag discrepancies for review—all while adapting to changes in the invoice portal’s layout without manual bot updates. Customer support teams can use the tool to scrape ticket details from a Zendesk view, feed them to a summarization LLM, and draft response suggestions that agents can approve with a single click.
Quantifying the impact of Pigeon Workflow involves looking at both direct cost savings and indirect benefits such as employee morale and agility. A mid‑size company that automates ten repetitive processes saving an average of five hours per week per process can reclaim roughly 2,600 hours annually—equivalent to 1.25 full‑time employees. Because the bots rarely break due to UI changes, the ongoing maintenance cost drops from an estimated 20 % of initial development effort per month to under 5 %. Moreover, by offloading tedious data‑entry and formatting tasks, staff report higher satisfaction and are able to devote more time to analytical work, problem solving, and customer engagement, which often translates into improved service quality and revenue growth. Early adopters have also noted a reduction in error rates: manual copy‑paste mistakes that once led to billing discrepancies dropped by over 80 % after automation, decreasing rework and associated compliance risks.
Successful adoption begins with a clear pilot. Organizations should select a process that is high volume, rule‑based but currently painful due to frequent UI changes or subjective judgment steps, and has measurable KPIs such as processing time or error rate. A cross‑functional team—comprising a business owner, an IT liaison, and a potential citizen developer—should receive a brief training on the Pigeon Workflow interface, then use the 14‑day free trial to build a prototype. During the pilot, it is essential to log both automation runtime and any manual interventions required; this data informs the ROI calculation and highlights any gaps in the AI node configurations. Governance is equally important: define who can publish workflows, establish version control, and set up monitoring alerts for failures or performance degradation. Once the pilot demonstrates a positive payback—typically within two to three months—the solution can be scaled to additional departments, guided by a center‑of‑excellence model that shares reusable nodes and best practices.
Looking ahead, Loftal plans to expand the library of pre‑built integration nodes for popular SaaS applications such as Salesforce, ServiceNow, HubSpot, and Zoom, allowing users to call APIs directly from within the visual flow without leaving the designer. Parallel to this, the company is investing in reinforcement‑learning techniques that observe a user’s repeated actions and suggest optimizations—such as combining two consecutive clicks into a single hover‑and‑select operation or recommending a more efficient LLM prompt. The ultimate vision is an “autonomous workflow assistant” that not only executes the designed steps but also proactively proposes new automations based on observed work patterns, continuously improving efficiency without manual intervention. This shift from reactive automation to proactive, AI‑driven process evolution could redefine how organizations view the role of digital workers, moving them from static scripts to adaptive partners in daily operations.
For readers evaluating whether Pigeon Workflow fits their needs, start by mapping out your most repetitive, browser‑based tasks and estimating the time lost to manual execution and error correction. Sign up for the free trial on the vendor’s website and build a simple workflow—perhaps automating a weekly report download and email distribution—to experience the visual recorder and AI nodes firsthand. Measure the baseline processing time, then compare it to the automated run after a week of use. If the results show a clear time saving and minimal maintenance overhead, consider expanding to a second, slightly more complex process that incorporates an LLM node for judgment. Keep a living document of lessons learned, share successful flows across teams, and establish a modest center‑of‑excellence to govern future deployments. By taking these incremental, evidence‑based steps, you can harness the power of AI‑driven no‑code automation to reduce drudgery, increase accuracy, and free your workforce to focus on higher‑value, creative work.