I asked ChatGPT Images 2.0 to redesign my app UIs – and wow


Can ChatGPT Images 2.0 improve your product's user interface?

David Gewirtz / Elyse Betters Picaro / ZDNET

Follow ZDNET: Add us as a preferred source on Google.


ZDNET’s key takeaways

  • I tested ChatGPT Images 2.0 on two active product UIs.
  • The AI found design issues and suggested practical fixes.
  • For solo developers, AI design review could be a big deal.

For the past week or so, I’ve been exploring the depths of OpenAI’s new ChatGPT Images 2.0 release. This is a much more consequential release than it first appears.

We’ve all been impressed with AI image generators. These tools make pretty pictures and can do some fun tricks. They can also produce a lot of AI slop. But, so far, they’ve been limited in their understanding of what they’re producing.

Also: 7 AI coding techniques I use to ship real, reliable products – fast

Images 2.0 adds subject-matter intelligence to image generation. This capability means ChatGPT can receive assignments and produce high-value outputs.

In this article, we’re going to look at one such type of assignment, updating a user interface. Previously, coding agents could take a shot at cleaning up a UI element here and there. The chatbots could make text-based recommendations on what needed fixing. But they couldn’t do the full design.

Now, they can. I fed two UI designs I’m actively working on into ChatGPT Images 2.0 on my $20/month ChatGPT Plus plan. In a few short minutes, ChatGPT returned two redesigned user interfaces, both of which resulted in a whole bunch of design improvements I plan to incorporate into my products.

Here are four recent articles that will bring you up to speed on what I’ve already found with this new tool:

I often use my non-ZDNET work as project fodder for ZDNET discovery articles. I try to avoid sharing product names and links because I don’t want any conflict of interest. But in this article, it’s necessary to show the product names. The names are part of the existing UI design that I’ve asked the AI to work on. If I pulled them out, it wouldn’t make as much sense. I’ll continue to avoid linking to the products themselves.

Also: I got 4 years of product development done in 4 days for $200, and I’m still stunned

Let’s get started.

Mac interface design

For this challenge, I had two interfaces I wanted to run by the AI. 

Also: I used Claude Code to vibe code a Mac app in 8 hours, but it was more work than magic

The first one is a Mac app I’ve been vibe coding with Claude Code since January. This project is taking a while because I only have an hour or two a week to work on it, and it uses MacOS’s internal AI for image processing and analysis.

Here’s the design as it exists in my app right now. A lot is happening here. I’m particularly invested in the big buttons on the left, because they reflect the brand colors for my wife’s e-commerce business. They also translate really well to the iPhone app:

stash-orig

Screenshot by David Gewirtz/ZDNET

I uploaded that screenshot to ChatGPT and fed it the prompt, “Redesign this user interface to make it more attractive and easier to use.” Here’s what ChatGPT Images 2.0 created:

stash-ai-annotated.png

David Gewirtz via ChatGPT Images/ZDNET

At first, I didn’t like the alterations. The most obvious change was the loss of the colored buttons. But the AI also didn’t quite understand that there are viewing options at the bottom of the grid view.

Also: How AI has suddenly become much more useful to open-source developers

On the other hand, look at those red squares. Those are the areas from the new mockup I really like, and plan to incorporate into my UI:

  • The set of actions on the lower left were icons on the iPhone app, but look far better with the UI presented by the AI.
  • Likewise, the AI added a much clearer header zone at the top of the grid view, as shown in the top middle red box.
  • The AI also spread out the pattern thumbnails, giving them more breathing room to make the selection box clear and attractive, as shown in the lower red box in the middle.
  • The AI added a favorites option that I haven’t wired in, but really like.
  • Finally, I like the idea of a persistent Added/Updated field at the bottom of the detail view.

My big takeaway from this exercise is those five new design notes I plan to incorporate into my product. I’m not redesigning the whole app to reflect the suggested UI. However, the image included some great ideas.

Plus, having a mockup like this will make it much easier to show Claude Code what I’d like it to do. This exercise is a great example of using the $20/month ChatGPT Plus plan and the $100/month Claude Code Max 5x plan to build better software.

Web interface design

Next up is the UI for the starting page for my security product. My design is clean but quite rudimentary, mostly reflecting the fact that I dislike coding in CSS.

Also: I did 24 days of coding in 12 hours with a $20 AI tool – but there’s one big pitfall

Last fall, in my first agent-based vibe coding project ever, I used OpenAI’s Codex in my ChatGPT Plus plan (which ran out of AI allotment quite fast) to redesign the UI from fairly ugly to unobjectionable. Here’s the result of that first pass last fall. I did the design, telling Codex where and how to lay out the CSS:

web-orig

Screenshot by David Gewirtz/ZDNET

The results from my testing this week were interesting. I pasted the screenshot into ChatGPT, but accidentally hit return before I could give it a prompt. ChatGPT decided on its own to analyze the page, deducing that “The biggest issues are weak visual hierarchy, too much gray, a very long intro block, and three lower cards that compete equally for attention even though they are not equally important.”

It also recommended a “more modern admin aesthetic,” including:

  • White card backgrounds instead of large gray panels.
  • Softer borders.
  • More spacing between sections.
  • One strong accent color, used consistently for buttons, links, and active states.
  • Shorter line lengths for body text.
  • Better contrast between headings, subheads, and body copy.

Also: Claude Code made an astonishing $1B in 6 months – and my own AI-coded iPhone app shows why

At this point, I decided to engage Images 2.0 and instructed, “Provide me an image of the redefined interface.” Here’s what the AI produced:

web-ai-annotated.png

David Gewirtz via ChatGPT Images/ZDNET

This output also provides some design notes I can go back to Codex with. I use OpenAI’s Codex with the web product and Claude Code with the Mac product, primarily to give me experience with the two agentic coding tools so I can write about them. If I didn’t have writing here as my primary goal, I’d choose one tool or the other. I’ve found both to be equally helpful.

Also: How Claude Code’s new auto mode prevents AI coding disasters – without slowing you down

ChatGPT Images 2.0 decided to invent a logo (mostly because I hadn’t provided it with one in my first image). The AI added several features I liked.

First, I really liked the overall design aesthetic it recommended. It might be a lot of work to propagate that through the rest of the product (it’s a giant product), but I did like the design.

Additionally, I liked how the AI promoted three separate user-help areas: the Quick Setup zone, the Need Help zone, and the Configure Privacy/View Docs section at the top. While the original interface gave users access to information, it really didn’t make it obvious where to begin.

I also liked the Site Status section at the bottom. I’ve often thought about adding a feature like that, but was a bit daunted by the coding requirements, because the status elements would have to change based on which freemium plugins the user had installed. Still, seeing a representation of that feature made me want to add it.

This is game-changing

I feel this capability is as game-changing as ChatGPT was back when it first came out, and as agentic pair programming was when it landed last summer.

Also: I built two apps with just my voice and a mouse – are IDEs already obsolete?

I submitted two product user interfaces to the AI and received essentially peer-reviewed commentary, along with a set of prototype designs. As a solo programmer, this output is invaluable.

Even if I had a full in-house team, with programmers and designers on staff, it would probably have taken a week or so to run this analysis and construct prototypes to review. The payroll expenses alone for that project would have been fairly substantial. But for $20, I was given two very helpful, very constructive, and not-at-all-AI-stupid redesigns that contained actionable nuggets that will make my products better.

Just wow.

If you were pairing ChatGPT Images 2.0 with Claude Code or Codex, would you use the image mockup as a design brief for the coding agent? Let us know in the comments below.


You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.





Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews


India’s financial sector is at a turning point. Gross NPAs of Scheduled Commercial Banks have fallen to a historic low of 2.15% as of September 2025, a figure not seen since 2010–11. Yet in absolute terms, gross NPAs still stand at approximately ₹4.32 lakh crore. The scale of the problem hasn’t disappeared; it’s shifted, from large corporate defaults to a more distributed mass of retail and MSME accounts scattered across geographies, legal jurisdictions, and ticket sizes.

For banks, NBFCs, and fintechs trying to recover these dues, understanding India’s debt recovery laws is not optional, it is foundational. This guide breaks down every major legal channel available, how they perform in practice, and what 2025’s regulatory shifts mean for lenders and recovery professionals.

At a Glance: India’s debt collection software market reached approximately $172.8 million in 2024 and is projected to reach $456 million by 2033 (CAGR of 10.48%, IMARC Group). Over 320 new debt recovery platforms launched between 2022 and 2024. The race is on, but legal infrastructure remains the backbone.

What Is Debt Recovery?

Debt recovery is the structured process by which lenders reclaim unpaid loan amounts from borrowers who have defaulted. Credit creation, through loans extended to individuals, MSMEs, and corporations, is essential to economic growth. But when borrowers default, lenders must navigate a complex web of legal mechanisms to recover what is owed. In India, this ecosystem spans eight distinct legal frameworks, multiple tribunals, and an increasingly digitised regulatory environment.

A loan account is classified as a Non-Performing Asset (NPA) when both principal and interest payments remain overdue for 90 days. Once classified as an NPA, lenders have access to several legal channels to recover dues, each with its own jurisdiction, timelines, and effectiveness.

Two Paths: Legal vs. Illegal Methods

The law draws a clear line between legitimate recovery and harassment. RBI guidelines require that all recovery communications occur strictly between 8 AM and 7 PM, agents carry valid identification, and no abusive or intimidatory tactics are used. The RBI’s February 2026 draft directions for both commercial banks and AIFIs (All India Financial Institutions) now mandate board-approved recovery policies, IIBF certification for agents, recording of recovery calls, and public disclosure of empanelled recovery agents, all effective July 1, 2026.

Illegal methods, public shaming, threats, late-night calls, or unauthorised property seizure, are not only unethical but expose lenders to regulatory action and grievances filed with the RBI Ombudsman. Nearly 39% of borrowers surveyed have reported abusive recovery calls; RBI data confirms that loan and credit-card complaints now form the largest single category of grievances received.

1. Indian Contract Act, 1872

Every loan relationship originates from a contract. If a borrower defaults, the lender can seek legal relief under several provisions of the Indian Contract Act, through a Contract of Guarantee (Section 126), Contract of Indemnity (Section 124), or by establishing Fraud (Section 17) or Misrepresentation (Section 18). This is typically a foundational step before more specific recovery mechanisms are invoked.

2. Civil Remedy (CPC Order IV)

A civil suit under Order IV of the Civil Procedure Code allows lenders to approach a court for money recovery. The suit must be filed within 3 years from the date of the cause of action and in the court that has jurisdiction over the borrower’s residence or place of business. Court fees are levied based on the claim amount. Civil suits are best suited for cases where other faster mechanisms are not available — but they are time-consuming and should be approached with a structured documentation trail.

3. Criminal Case Under IPC (Now BNS, 2023)

Where the default involves elements of cheating, criminal breach of trust, or dishonest misappropriation, lenders can file a criminal case. Key provisions include Cheating (Sections 415/417 IPC, now mirrored in the Bharatiya Nyaya Sanhita, 2023), Criminal Breach of Trust (Sections 405/406), and Dishonest Misappropriation of Property (Section 403). Some of these offences are non-bailable and cognizable, meaning the defaulter faces serious legal consequences.

4. Insolvency and Bankruptcy Code (IBC), 2016

The IBC remains India’s most powerful corporate debt recovery instrument. Where the defaulted amount exceeds ₹1 crore (revised from ₹1 lakh in 2020), creditors can approach the NCLT for initiating the Corporate Insolvency Resolution Process (CIRP). A Committee of Creditors (CoC) is formed, an Insolvency Professional appointed, and the resolution must be approved by 66% of CoC votes within 330 days.

IBC Impact by the Numbers (as of March 2025):
— Over 30,000 applications involving defaults of ₹13.78 lakh crore were settled at the pre-admission stage alone, demonstrating IBC’s deterrence effect.
— Average recovery rates improved from 15–20% pre-IBC to approximately 30% post-IBC (S&P Global Ratings, December 2025).
— S&P upgraded India’s insolvency regime from ‘Group C’ to ‘Group B’ in December 2025.
— However, actual average CIRP duration stands at 713 days, more than double the statutory 330-day limit. NCLT pendency is nearly 30,600 cases (March 2025), with an estimated 10-year clearance time at current rates.

IBC’s biggest strength is its behavioural impact, it has fundamentally shifted the culture from “debtor in possession” to “creditor in control.” The proportion of overdue corporate loan amounts relative to total outstanding fell from 18% in 2018 to 9% in 2024 (IIM Bangalore study).

5. Negotiable Instruments Act, Section 138 (Cheque Bounce)

One of the most frequently invoked debt recovery provisions in India, Section 138 of the NI Act applies when a post-dated or security cheque issued by a borrower is returned unpaid. Upon dishonour, the payee must send a demand notice within 30 days; if the borrower fails to make payment within 15 days, criminal proceedings can be initiated. The defaulter may face imprisonment of up to 2 years, a fine twice the cheque amount, or both. Cheque bounce cases number in the millions annually across Indian courts, making efficient case management critical for lenders handling high volumes.

6. RDDBFI Act, 1993, Debt Recovery Tribunals (DRTs)

The Recovery of Debts Due to Banks and Financial Institutions Act established a network of 39 Debt Recovery Tribunals (DRTs) and 5 Debt Recovery Appellate Tribunals (DRATs) across India. Banks and NBFCs can file applications under Section 19 for recovery of dues. Borrowers who wish to appeal a DRT order must deposit 50% of the debt amount (reducible to 25% by the appellate tribunal). While DRTs were designed for speed, chronic understaffing and high pendency have limited their effectiveness. DRTs accounted for just 4.2–4.9% of total NPA recovery in recent years, among the lowest of all channels.

Note on DRT Reform: The government has signalled intent to expand DRT jurisdiction and address vacancies. The BAANKNET e-auction portal, launched March 25, 2025, is already improving asset disposal efficiency for PSBs and IBBI-referred cases.

7. SARFAESI Act, 2002

The Securitisation and Reconstruction of Financial Assets and Enforcement of Security Interest Act allows secured creditors, banks, NBFCs, and ARCs, to take possession of and sell secured assets without court intervention. Once a loan is classified as NPA under Section 13, a notice is sent to the defaulter giving 60 days to repay. If repayment doesn’t happen, the lender can sell the asset or assign it to an Asset Reconstruction Company (ARC) at a discounted rate.

SARFAESI is particularly favoured by banks due to lender control over the asset sale process. It accounted for 17.4–26.7% of total NPA recovery in recent reported years. Recent amendments have strengthened the framework further, including empowering RBI to audit ARCs and mandating CERSAI registration of security interests.

8. Summary Suit

A Summary Suit (Order XXXVII, CPC) is a fast-track civil proceeding suited for liquid debts not exceeding ₹10 lakh. The defaulter has just 10 days from the date of service to appear before the court. If they fail to do so, the court may pass an ex-parte decree immediately. While the ticket-size cap limits its use for large institutional lending, it is a practical tool for smaller NBFC or retail exposures.

How Each Channel Actually Performs: Recovery Rate Comparison

Recovery Channel Share of Recovery (Recent Years) Average Timeline Best Suited For
IBC / NCLT ~44–46% (highest among all channels) 713 days average (statutory: 330 days) Large corporate defaults >₹1 crore
SARFAESI Act 17–27% Months (no court required) Secured assets, banks & larger NBFCs
DRTs 4.2–4.9% 1–3+ years (due to pendency) Mid-size bank/FI claims
Lok Adalats ~6% (low recovery per case) Weeks to months Small-ticket pre-NPA settlements
Section 138 / NI Act Varies (high volume, lower value) 1–3 years in metro courts Cheque-secured loans
Civil Suits Varies 3–7 years Unsecured creditors, contractual disputes

Sources: RBI Annual Reports, IBBI data, Lexology analysis, IBC Laws research platform, FACTLY data analysis (March 2025).

RBI’s 2025–26 Guidelines: What’s Changing for Lenders

The regulatory landscape for debt recovery shifted significantly in 2025. Three key developments stand out:

1. RBI Digital Lending Directions, 2025 (effective May 8, 2025) — This consolidated framework governs all digital lending activity including recovery. Lenders must notify borrowers via email/SMS before any recovery agent makes contact, ensure all disbursals go directly to borrower bank accounts, and maintain transparent grievance channels. Lending Service Providers (LSPs) acting as recovery agents are now held to the same standards as the Regulated Entity (RE) itself.

2. Draft Responsible Business Conduct (Amendment) Directions, February 2026 — Released simultaneously for commercial banks and AIFIs, these draft directions (effective July 1, 2026) represent the most comprehensive overhaul of recovery conduct standards in years. Key mandates include: board-approved recovery policy, IIBF certification for all recovery agents, mandatory recording of recovery calls, public disclosure of empanelled agents, written notice of default before any recovery action, and strict prohibition on harsh practices including public shaming, abusive language, and family/colleague harassment.

3. BAANKNET Portal, March 2025 — The government’s revamped e-auction platform integrates all 12 Public Sector Banks and IBBI with automated KYC, secure payments, and bank-verified property titles, significantly improving transparency in SARFAESI-based asset sales.

Compliance Implication for Lenders: Legal recovery today is increasingly about process documentation, not just legal filing. A timestamped, digitally-traceable record of every notice, communication, and action is no longer just operationally helpful — it is a regulatory requirement. A WhatsApp chat archive will not hold up under RBI or DRT scrutiny.

Best Practices for Lenders Navigating the Legal System

Build a Structured Internal Process Before Filing

Debt recovery requires coordination across internal legal, finance, and collections teams — and often, an external advocate or law firm. Designate clear accountability: who signs the notice, who coordinates with external counsel, who monitors hearing dates. Manual calendar-based tracking of court dates leads to adjournments, value erosion, and missed opportunities. Automated case management — with alerts triggered by hearing schedules, advocate assignments, and SLA breaches — is the baseline for any serious recovery operation today.

Document Everything, Digitally

Every communication with the borrower — from the first demand notice to field visit reports — must be documented with timestamps. This is not just good practice; it directly affects your legal standing. In SARFAESI and DRT proceedings, the quality and completeness of the paper trail often determines outcomes. Automated notice dispatch that generates a delivery-confirmed, timestamped audit log gives lenders a defensible record.

Choose the Right Jurisdiction Before Filing

Filing in the wrong court or tribunal is a costly, time-consuming error. Match the legal channel to the debt type and ticket size: IBC/NCLT for large corporates (>₹1 crore), SARFAESI for secured assets, DRT for bank/FI claims, Section 138 for cheque bounce, civil suits or Lok Adalats for smaller unsecured accounts. For retail and MSME NPA accounts with smaller ticket sizes, pre-litigation ODR (Online Dispute Resolution) platforms are emerging as a cost-effective alternative to formal proceedings.

Engage Qualified Counsel, and Track Their Performance

Advocate selection in recovery litigation is frequently based on familiarity rather than performance data. This leads to systemic underperformance. High-performing lenders are increasingly using data to track advocate win rates, adjournment frequency, and case resolution timelines by jurisdiction, and adjusting their panels accordingly.

Maintain Ethical Standards to Protect Your Recovery

Courts and tribunals look at the conduct of both parties. A lender that can demonstrate ethical, documented, and RBI-compliant recovery behaviour before filing is better positioned to receive favourable outcomes. Violations of RBI conduct guidelines, even if not the direct subject of the case, can undermine a lender’s standing.

The Role of Technology in Modern Debt Recovery

The 2024–25 period has seen a structural shift in how lenders approach recovery infrastructure. AI is now deployed across predictive default scoring, omnichannel borrower communication, automated legal notice dispatch, and court case management. Mid-sized banks have reported a 34–36% reduction in collection costs after AI adoption, with recovery rate improvements of 10–25%.

The most significant strategic shift is toward ecosystem thinking rather than monolithic platform adoption. Different parts of the recovery journey require different tools: pre-litigation communication platforms for early-stage accounts, ODR/mediation for small-ticket disputes, and dedicated legal operations infrastructure for NPA accounts heading to DRT, SARFAESI, or NCLT. The bridge between collections-stage activity and legal-stage activity, where cases are handed off, documents compiled, and notices issued, remains the most operationally fragile point in most lenders’ recovery chains.

Key Technology Stats for Recovery Professionals:
— AI adoption in mid-size banks: 34–36% cost reduction in collections
— Recovery rate improvement post-AI: 10–25%
— India’s debt collection software market CAGR: 10.48% (2024–2033)
— PSB gross NPA ratio: 2.50% (September 2025)
— Private sector bank NPA ratio: 1.73% (September 2025)

The Bottom Line

India’s debt recovery legal framework is comprehensive, and under active improvement. The IBC has reshaped creditor rights. SARFAESI gives secured lenders direct enforcement power. The 2025–26 RBI guidelines are tightening conduct standards while pushing for digital accountability. And the absolute scale of NPAs, despite improving ratios, means the demand for effective, tech-enabled, legally defensible recovery will only grow.

For lenders, the question is no longer whether to digitise their legal recovery operations, but how quickly they can build infrastructure that is compliant, data-driven, and defensible at every stage, from first notice to final court order.


Want to see how Legodesk connects your collections workflow directly to legal recovery, from automated notice dispatch to court case management, notice tracking, and recovery through Lok adalat? Request a demo



Source link