Posts by Date – Israeli AI Community
11.2
Important update from the past week: Claude Opus 4.6 has been launched, and it brings a significant change in how the model handles complex tasks.
The central innovation in this version is a mechanism called Adaptive Thinking. The practical meaning: the model knows how to regulate its thinking depth according to the task. For simple tasks, it responds quickly, and for complex tasks, it invests more time and processing, without the user having to define it beforehand.
In previous generations, two edge case issues were sometimes encountered: answers that were too superficial, or conversely, getting bogged down in calculations and analyses longer than necessary. In 4.6, Anthropic attempts to automatically balance speed and accuracy, resulting in a noticeable improvement in the stability and quality of answers, especially in business, legal, and technological analyses.
One of the uses where the difference is particularly noticeable is working with complex documents. Instead of guiding the model step by step on what to look for, you can upload a broad file and ask it to formulate a direction, analysis, or strategic framework, with the model itself employing a multi-stage thinking process to reach the result. The model is also available to users from Israel, including free access with a limited quota, alongside advanced paid plans.
4.2
Did you know? This is how you turn text into a visual chart without leaving the chat.
Imagine being able to turn a verbal description into a professional visual diagram without leaving the conversation window for even a moment. The direct connection between ChatGPT and Figma allows precisely that: using the "@" symbol connects the chat to external interfaces and generates visual outputs for you in seconds.
How does it work? Instead of struggling with manual design, you simply type "@figma" in the chat and add a simple request: "Create me a flowchart of a new client onboarding process." The model activates Figma's capabilities and generates, right before your eyes and within the conversation itself, a perfect diagram ready to be included in your presentation or workflow. This is not just a time saver, it's a shortcut that turns complex workflows into a clear and accessible product at the click of a button.
💡 Important tip: For the magic to happen, make sure that in your ChatGPT settings, the connection to Figma is enabled.
27.1
Remember the hours we spent cracking complex Excel formulas? Today, writing formulas has become a simple task thanks to a verbal description of the professional need.
Instead of getting complicated with multiple condition functions, simply write your request to the model as is: "I have an accounts receivable report as of 12/31/2025. Create a new analysis column: if the debt in column D is 3-4 years old, list 'Provision for Doubtful Accounts'; if it is over 5 years old, list 'Bad Debt'; and if it is under 3 years old, leave it blank. Do not change the original data, and return an updated Excel file with the analysis.".
The model not only generates the formula (e.g., a nested IF function), but can also process the file for you and return it ready for download with the new analysis columns, without touching the source data. This is an exceptionally effective solution for building management reports that combine data from various sources without requiring prior coding knowledge.
Little tip: To ensure the analysis is accurate on the first try, it's best to specify to the model which column and row the relevant data is located in (e.g., "The balances are in column D starting from row 2").
Just think about a function you're used to putting a lot of effort into in Excel, and describe it in your own words. This is the best way to practice writing an accurate prompt and achieve accurate and fast results.
19.1
How to improve accuracy when working with language models?
Simple example: When requesting a complex social security calculation, it is advisable to add the phrase "Solve step-by-step." Instead of receiving a final number that might be incorrect, the model shows all the arithmetic calculation steps. This allows you to review the logic and ensure the result is correct. Another tip is to define in the model's permanent instructions: "Always present answers in a table with a source column and a conclusion column." Such a definition saves the need to request data reordering for every question and ensures consistent, organized, and easy-to-read information every workday.
12.1
The combination of Gamma and Nano Banana saves hours of graphic work. Instead of manually designing a presentation for the finance department, you type a command into Gamma: "Create a presentation on tax regulation changes." The system builds a structure and slides within minutes. Nano Banana produces professional background images that match the company's brand.
Practical example: Upload a Word file with cash flow data to Gamma, and it automatically turns it into a structured, ten-slide presentation. The speed allows you to focus on analyzing the conclusions instead of arranging boxes and colors on the screen. The ability to produce high-level visual materials with minimal effort is a significant advantage when presenting data to management.
30.12
The next revolution in accounting: from historical analysis to strategic forecasting (Predictive AI)
The year 2026 marks the crucial transition from historical data analysis to artificial intelligence in forecasting.
This is a fundamental change: using existing information to predict trends and assess future risks in real-time. The structural change transforms the accountant from a passive data reporter into a strategic advisor who safely guides the business.
Forecasting capabilities enable high-accuracy cash flow generation by analyzing macro variables and specific payment patterns. Additionally, AI systems allow for continuous monitoring and real-time anomaly detection, replacing traditional statistical sampling methods in auditing. This type of technology also enables the running of complex "What-if Analysis" scenarios, which aid in optimal tax planning and investment feasibility assessment.
To implement these capabilities in your office, it is recommended to consider using dedicated tools: DataRails, which enables automation of reports and advanced financial forecasting based on Excel data; Power BI in conjunction with Copilot components for visual trend analysis; and platforms like Jirav, which specialize in AI-based automation and operational and cash flow forecasting.
Adopting these tools provides value that benefits the client far beyond meeting dry regulatory requirements. The shift to proactive accounting strengthens the economic resilience of the firm and its clients, ensuring a competitive edge in a dynamic market.
23.12
Meet Google Workspace Studio: Build AI agents that work for you within your email and Excel.
Friends, Google launched Workspace Studio this month, and it's significant news for anyone running a business, office, or team. This is a revolutionary platform that allows each of us to build a sophisticated personal "AI agent" using a no-code method – meaning, without needing to write a single line of code. What makes this tool unique is its ability to function as an autonomous connecting thread between the applications we all work with daily: Gmail, Drive, Sheets, and Calendar.
How does this process upgrade office and business management?
End-to-end lead and inquiry management: Instead of manual data entry, you can set up an agent that identifies new inquiries in email, automatically extracts contact details and requirements, updates the tracking sheet in Sheets, and sends the client a follow-up message or a relevant price quote based on existing files in Drive.
Smart diary and meeting content management: The agent can independently manage dialogue with external parties to find available and suitable times in the diary. Additionally, it can analyze meeting transcripts, extract action items from the content, and send them to all participants immediately after the conversation ends.
Automating Document Filing and Classification: Imagine an agent that scans every invoice, contract, or legal document that arrives via email, analyzes who the client is and the document type, and automatically files it into the correct cloud folder while updating the status in your control table. This capability allows us to stop wasting time on technical tasks and start managing a team of digital agents that work for us around the clock with peak precision and speed.
18.12
Meet Gemini 3 Flash: The World's Fastest Engine for Financial Data Analysis
Friends, colleagues, Google's new Gemini 3 Flash model redefines the speed at which massive amounts of information can be processed. The model's uniqueness lies in its context window of up to one million tokens, allowing the system to ingest dozens of financial reports or hundreds of Excel spreadsheets simultaneously and receive cross-sectional analysis in just a few seconds.
How does the process help with financial management?
Contract scans and exposure identification: Upload dozens of supplier contracts at once to automatically locate outlier indexation clauses or renewal dates. The model can compare all contracts simultaneously.
Customer and Supplier Analysis: Cross-referencing procurement and sales data to identify concentration, payment morality, and supply chain trends.
Multi-year cash flow analysis: Reading a long history of bank statements and profit and loss statements to identify seasonal trends and deviations without needing to split the information into smaller files.
Important privacy emphasis: Avoid uploading identifying details such as company registration numbers or identifiable client names to maintain organizational data security.
15.12
Friends and colleagues, as finance managers and business owners, you're accustomed to looking at AI as a tool for data analysis or writing emails. But Google's new model, Gemini 3, allows for dynamic and flexible work – and that makes it a strategic tool, not just a technical one.
While other tools give answers "by the book," this model knows how to step into the shoes of different characters and perform complex simulations. Here's how it looks compared to the competition: While ChatGPT excels at cold logic and Claude is a champion at analyzing long documents, Gemini 3 is the "creative" one of the bunch. It knows how to break out of the standard mold, change tone in real-time, and offer non-linear solutions.
How does it serve you in your day-to-day life? Here are two examples that change the game:
The "War Room": A Negotiation Simulation. Before you enter a critical meeting with the bank to increase your credit line or with a stubborn supplier, the model serves as a practice partner. You don't ask it to "write arguments," but rather tell it: "You are now the bank's risk manager, be tough and challenge me." It will conduct a live dialogue with you, attack the weak points in your cash flow, and help you arrive at the meeting prepared for any scenario.
Storytelling for the Board and Investors: Presenting an Excel sheet with an annual forecast is easy. Telling the story behind the numbers – that's the challenge. The model's dynamic capability knows how to take the same dry data and generate 3 different narratives from it at the press of a button: one conservative and reassuring version for the bank, a second aggressive version for venture capital investors, and a third marketing version for employees. It understands that the number is the same number, but the story must change according to the target audience.
10.12
Friends, after the initial excitement, we tested Google's new model (based on Gemini 3) to understand if it's a gimmick or a tool. The conclusion? The new toolkit solves most of the problems that have been frustrating us until now with AI. Let's start with Style Reference, the feature we've been waiting for. Do you have a picture with perfect lighting or a unique brand illustration style? You can upload it as a "reference." The model will not copy the content, but rather "learn" the style (color palette, mood, texture) and apply it to any new idea you have. This is a game-changer for maintaining a consistent design language.
Surgical editing (Inpainting) - Did you create an amazing image but there's an unnecessary potted plant in the background? A hand turned out weird? No need to regenerate and hope for the best. You can mark a specific area and ask the model to fix only that. Changing a suit to a t-shirt or removing an object – it all happens in seconds without compromising the composition.
The days of models getting "lost" with complex prompts are over. The "Nano Banana" demonstrates impressive spatial understanding: ask for "three people in a circle, the one on the right holding an iPad and the one on the left pointing to the sky" – and it will place them exactly there. Less trial and error, more results.
Character Consistency. The big news: the ability to generate a character (presenter) and run with it through a whole series of images in different situations, thanks to the ability to remember up to 14 reference images.
Hebrew without gibberish – The model knows how to integrate text into images, in various fonts, and in full, standard Hebrew. Signs, invitations, and banners.
1.12
Google breaks the tools with "Nano Banana Pro" – what are the must-know innovations?
Friends, if until today we were accustomed to image models being nice but limited, things are changing starting today. Google launched "Nano Banana Pro" (based on Gemini 3) this week and proved that it is aiming for genuine professional use.
It's no longer just about "create me a cute picture," but a powerful tool for creators and businesses. We checked out the innovations worth knowing about: The end of "changing faces" (Consistency) – perhaps the biggest news: the model allows uploading up to 14 reference images and maintaining consistency for up to 5 different characters. What does this mean? You can produce a series of images, comics, or a campaign with the exact same presenter in different situations. Full Hebrew and text design: The days of gibberish are over. The new model supports integrating texts into images in a variety of fonts, and yes – also in Hebrew. The model opens the door to creating ready-to-publish invitations, posters, and marketing materials. Studio photographer control: Google has introduced upgraded editing capabilities (In-painting). Didn't like the lighting? Want to change the camera angle or focus? You can edit specific elements within the image without regenerating it.
From "napkin sketch" to business graph: A powerful presentation feature – The model knows how to take scribbles of points you wrote on a note, or raw data, and turn them into a visually impressive, well-designed graph.
28.11
What did we learn this week from hundreds of accountants?
Friends and colleagues, what a week it's been – on Monday, we launched a first-of-its-kind course in collaboration with the Hebrew University to enrich accountants in the field of AI. On Wednesday, we met for a special AI applications webinar.
Both events were initiated by the managing partner of Barzilai & Co. CPA, Daphna Ravid Barzilai, who also serves as the Vice President of the Institute of Certified Public Accountants in Israel and Chair of the AI Committee. The meetings proved once again that the thirst for knowledge is immense, but most importantly, that we have collectively moved past a stage: we are no longer just talking about theory, but about practical work.
In meetings, Dafna and other speakers presented how AI performs magic in complex simulations, such as preparing a financial report analysis complete with charts and presentations, but at the same time, we understood exactly where its limitations lie. It knows how to process data, but it doesn't know how to "read the room" or understand deep business strategy. And this is the reassuring news: accuracy comes from the machine, but wisdom, context, and direction? They only come from us, the professionals.
