Welcome input on my AI-powered monthly investment review workflow

I'm a busy professional who always wanted to spend more time managing my stock portfolio but never had the structure or time to do it well. As part of my experimentation with AI agents, I built a monthly review workflow that runs automatically on the 15th. Here's the sequence:

**1. Macro backdrop first** : scores 6 cross-asset ratios (RSP/SPY, yield curve, HYG/LQD, IWM/SPY, SPY/TLT, XLY/XLP) to identify the current market regime before touching individual stocks. If it reads contraction, everything downstream defaults to defense mode.

**2. Watchlist scan** : two screens run automatically: undervalued candidates (low forward P/E vs sector, FCF yield above 3%, insider buying) and oversold pullback candidates (down 20-40% from highs with intact fundamentals). The logic tries to filter broken stocks from genuine pullbacks.

**3. Per-position review** : for each current holding: fundamental trend, thesis health (intact / wobbling / broken), technicals, then a clean ADD / HOLD / WATCH / TRIM / EXIT with a reason attached.

**4. Institutional flow check** : for any ADD or EXIT candidate, cross-checks 13F filings to see if smart money agrees. Has a 45-day data lag by design, so I treat it as a confirming signal rather than a leading one.

**5. Exposure calibration** : synthesizes everything into a recommended equity allocation ceiling and whether new positions are cleared for that month.

Built for a 6-15 stock focused portfolio, medium-to-long hold periods, medium risk. Likely not useful for day traders or deep value investors with a lot of time for manual analysis.

What would you change or add? Particularly curious whether the macro regime step adds real signal or just noise, and whether anyone runs a similar workflow today, manually or otherwise.